Author: "Traitement du Langage Parlé (TLP)" / Search Limiters: Available in Library Collection - Searchworks@Jio Institute Digital Library Search Results

Your search keyword '"Traitement du Langage Parlé (TLP)"' showing total 57 results

Start Over Author "Traitement du Langage Parlé (TLP)" Search Limiters Available in Library Collection

57 results on '"Traitement du Langage Parlé (TLP)"'

1. Multimodal Inverse Cloze Task for Knowledge-based Visual Question Answering

Author: Lerner, Paul, Ferret, Olivier, Guinaudeau, Camille, Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Traitement du Langage Parlé (TLP ), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Département Intelligence Ambiante et Systèmes Interactifs (DIASI), Laboratoire d'Intégration des Systèmes et des Technologies (LIST (CEA)), Direction de Recherche Technologique (CEA) (DRT (CEA)), Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Direction de Recherche Technologique (CEA) (DRT (CEA)), Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Commissariat à l'énergie atomique et aux énergies alternatives (CEA)-Université Paris-Saclay, and ANR-19-CE23-0028,MEERQAT,Représentation multimédia d'entités et systèmes de question réponse(2019)
Subjects: FOS: Computer and information sciences, Computer Science - Machine Learning, Computer Science - Computation and Language, [INFO.INFO-MM]Computer Science [cs]/Multimedia [cs.MM], Multimodal Fusion, Visual Question Answering, Machine Learning (cs.LG), Multimedia (cs.MM), Computer Science - Information Retrieval, Pre-training, [STAT.ML]Statistics [stat]/Machine Learning [stat.ML], [INFO.INFO-IR]Computer Science [cs]/Information Retrieval [cs.IR], Computation and Language (cs.CL), Information Retrieval (cs.IR), Computer Science - Multimedia
Abstract: We present a new pre-training method, Multimodal Inverse Cloze Task, for Knowledge-based Visual Question Answering about named Entities (KVQAE). KVQAE is a recently introduced task that consists in answering questions about named entities grounded in a visual context using a Knowledge Base. Therefore, the interaction between the modalities is paramount to retrieve information and must be captured with complex fusion models. As these models require a lot of training data, we design this pre-training task from existing work in textual Question Answering. It consists in considering a sentence as a pseudo-question and its context as a pseudo-relevant passage and is extended by considering images near texts in multimodal documents. Our method is applicable to different neural network architectures and leads to a 9% relative-MRR and 15% relative-F1 gain for retrieval and reading comprehension, respectively, over a no-pre-training baseline., Accepted at ECIR 2023
Published: 2023

2. Language Report French

Author: Adda, Gilles, Vasilescu, Ioana, Yvon, François, Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Georg Rehm, Andy Way, and European Project: LC-01641480,ELE
Subjects: Language Resources, French Language, Human Language Technologies, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL]
Abstract: International audience; This chapter presents a survey of the current state of technologies for the automatic processing of the French language. It is based on a thorough analysis of existing tools and resources for French, and also provides an accurate presentation of the domain and its main stakeholders (Adda et al. 2022). The chapter documents the presence of French on the internet and describes in broad terms the existing technologies for the French language. It also spells out general conclusions and formulates recommendations for progress towards deep language understanding for French.; Ce chapitre présente une étude de l'état actuel des technologies pour le traitement automatique de la langue française. Il s'appuie sur une analyse approfondie des outils et ressources existants pour le français et fournit également une présentation précise du domaine et de ses principaux acteurs (Adda et al. 2022). Le chapitre documente la présence du français sur l'internet et décrit en termes généraux les technologies existantes pour la langue française. Il énonce également des conclusions générales et formule des recommandations pour progresser vers la compréhension approfondie de la langue française.
Published: 2023

3. Investigating the Translation Performance of a Large Multilingual Language Model: the Case of BLOOM

Author: Bawden, Rachel, Yvon, François, Automatic Language Modelling and ANAlysis & Computational Humanities (ALMAnaCH), Inria de Paris, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria), Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), and ANR-19-P3IA-0001,PRAIRIE,PaRis Artificial Intelligence Research InstitutE(2019)
Subjects: FOS: Computer and information sciences, Computer Science - Computation and Language, MT, NMT, LLMs, Machine translation, Large language models, Evaluation, Computation and Language (cs.CL), [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL]
Abstract: The NLP community recently saw the release of a new large open-access multilingual language model, BLOOM (BigScience et al., 2022) covering 46 languages. We focus on BLOOM's multilingual ability by evaluating its machine translation performance across several datasets (WMT, Flores-101 and DiaBLa) and language pairs (high- and low-resourced). Our results show that 0-shot performance suffers from overgeneration and generating in the wrong language, but this is greatly improved in the few-shot setting, with very good results for a number of language pairs. We study several aspects including prompt design, model sizes, cross-lingual transfer and the use of discursive context., Comment: Accepted at EAMT 2023
Published: 2023
Full Text: View/download PDF

4. Graph-Based Multilingual Label Propagation for Low-Resource Part-of-Speech Tagging

Author: Imani, Ayyoob, Severini, Silvia, Sabet, Masoud, Yvon, François, Schütze, Hinrich, Ludwig Maximilian University [Munich] (LMU), Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), and Association for Computational Linguistics
Subjects: [INFO.INFO-TT]Computer Science [cs]/Document and Text Processing, FOS: Computer and information sciences, Computer Science - Computation and Language, Transfert Cross-lingue, Cross-lingual Transfer, PoS-tagging, Etiquetage en parties du discours, Traitement automatique des langues, Computation and Language (cs.CL), Natural Language Processing
Abstract: Part-of-Speech (POS) tagging is an important component of the NLP pipeline, but many low-resource languages lack labeled data for training. An established method for training a POS tagger in such a scenario is to create a labeled training set by transferring from high-resource languages. In this paper, we propose a novel method for transferring labels from multiple high-resource source to low-resource target languages. We formalize POS tag projection as graph-based label propagation. Given translations of a sentence in multiple languages, we create a graph with words as nodes and alignment links as edges by aligning words for all language pairs. We then propagate node labels from source to target using a Graph Neural Network augmented with transformer layers. We show that our propagation creates training sets that allow us to train POS taggers for a diverse set of languages. When combined with enhanced contextualized embeddings, our method achieves a new state-of-the-art for unsupervised POS tagging of low-resource languages., EMNLP 2022
Published: 2022

5. Integrating Translation Memories into Non-Autoregressive Machine Translation

Author: Xu, Jitao, Crego, Josep, Yvon, François, Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), SYSTRAN, and PhD Talent Région Ile de France
Subjects: FOS: Computer and information sciences, Computer Science - Computation and Language, Computation and Language (cs.CL), [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL]
Abstract: Non-autoregressive machine translation (NAT) has recently made great progress. However, most works to date have focused on standard translation tasks, even though some edit-based NAT models, such as the Levenshtein Transformer (LevT), seem well suited to translate with a Translation Memory (TM). This is the scenario considered here. We first analyze the vanilla LevT model and explain why it does not do well in this setting. We then propose a new variant, TM-LevT, and show how to effectively train this model. By modifying the data presentation and introducing an extra deletion operation, we obtain performance that are on par with an autoregressive approach, while reducing the decoding load. We also show that incorporating TMs during training dispenses to use knowledge distillation, a well-known trick used to mitigate the multimodality issue., Accepted at EACL 2023 main conference
Published: 2022

6. Syntaxe théorique et formelle: Volume 1: Modélisation, unités, structures

Author: Kahane, Sylvain, Gerdes, Kim, Modèles, Dynamiques, Corpus (MoDyCo), Université Paris Nanterre (UPN)-Centre National de la Recherche Scientifique (CNRS), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Traitement du Langage Parlé (TLP ), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Kahane, Sylvain, and Gerdes, Kim
Subjects: 400 Sprache::410 Linguistik::410 Linguistik, [SHS.LANGUE]Humanities and Social Sciences/Linguistics, Syntaxe théorique et formelle
Abstract: International audience; Conceived as a general introduction to syntax, this book presents the basic concepts necessary for a study of the combination of lexical and grammatical units within an utterance. The book does not impose a preconceived framework, but rather examines the various possibilities for the representation of syntactic structures, according to the general principles and specific criteria that have been adopted.The aim of this book is to provide a basis for teaching syntax at university, and to show that it is possible to identify the properties of languages in a methodical way and to put order in the jungle of each language. The book is divided into three parts: How to develop the model of a language? How to determine the basic units of a language according to their meaning, form, and combinatorial nature? How to define and represent the different ways in which the units are organized? This last part presents an abundance of syntactic diagrams of a wide range of types.The book is divided into small sections, alternating the main content with insights, historical notes, formal elaborations, linguistic examples in diverse languages, proposals for further reading, and exercises with answer keys.; Conçu comme une introduction générale à la syntaxe, cet ouvrage présente les notions de base nécessaires à une étude de la combinaison des unités lexicales et grammaticales au sein d’un énoncé. Sans se placer dans un cadre préconçu, l’ouvrage étudie les différentes possibilités pour la représentation des structures syntaxiques, en fonction des principes généraux et des critères particuliers retenus.Élaboré avec l’objectif de fournir une base pour l’enseignement de la syntaxe à l’université, cet ouvrage souhaite montrer qu’on peut dégager de manière méthodique les propriétés des langues et mettre de l’ordre dans la forêt vierge que constitue chaque langue. Il est divisé en trois parties : comment élaborer le modèle d’une langue, comment déterminer les unités de base de la langue en fonction de leur sens, forme et combinatoire, comment définir et représenter les différents modes d’organisation des unités. Cette dernière partie présente une abondance de diagrammes syntaxiques de diverses natures.L’ouvrage est découpé en de petites sections, alternant le contenu principal avec des éclairages, des notes historiques, des élaborations plus formelles, des exemples linguistiques dans diverses langues, des propositions de lectures additionnelles et des exercices avec des éléments de correction.
Published: 2022

7. Modeling the prosodic forms of Discourse Markers

Author: Raso, Tommaso, Rilliard, Albert, Mendes Santos, Saulo, Universidade Federal de Minas Gerais = Federal University of Minas Gerais [Belo Horizonte, Brazil] (UFMG), Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), and Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)
Subjects: Spontaneous Speech, Fala Espontânea, Modeling, Discourse Markers, Prosody, [SHS.LANGUE]Humanities and Social Sciences/Linguistics, Prosódia, Modelização, Marcador Discursivo
Abstract: This paper has a twofold goal: (i) to propose how and why to identify Discourse Markers (DM), showing that the formal features marking this category are of prosodic nature and can distinguish the six different functions of interactional nature performed by DMs. We describe both the prosodic characteristics responsible for a DM identification and the prosodic forms that convey each type of communicative function inside the more general category of DM; (ii) to show in detail the methodological steps adopted so far to allow the automatic extraction of different DMs from new data. The methodology is presented together with a statistical-computational discussion and explanation. Este artigo tem um objetivo duplo: (i) avançar uma proposta para a identificação da categoria de Marcador Discursivo (MD), mostrando que as marcas formais do MD são de natureza prosódica e também capazes de distinguir cerca de seis diferentes funções de natureza interacional veiculadas pelos MDs. Se descrevem tanto as características prosódicas responsáveis para a identificação de um MD quanto as formas prosódicas que veiculam cada tipo de função comunicativa dentro da categoria maior de MD; (ii) mostrar detalhes da metodologia que em maior medida será adotada para modelizar essas unidades e permitir uma extração automática a partir de novos dados. Ela é apresentada com uma reflexão estatístico computacional que a justifica.
Published: 2022

8. Joint Generation of Captions and Subtitles with Dual Decoding

Author: Xu, Jitao, Buet, François, Crego, Josep, Elise Bertin-Lemée, Yvon, François, Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), SYSTRAN, and Xu, Jitao
Subjects: FOS: Computer and information sciences, Computer Science - Computation and Language, [INFO.INFO-CL] Computer Science [cs]/Computation and Language [cs.CL], Computation and Language (cs.CL), [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL]
Abstract: As the amount of audio-visual content increases, the need to develop automatic captioning and subtitling solutions to match the expectations of a growing international audience appears as the only viable way to boost throughput and lower the related post-production costs. Automatic captioning and subtitling often need to be tightly intertwined to achieve an appropriate level of consistency and synchronization with each other and with the video signal. In this work, we assess a dual decoding scheme to achieve a strong coupling between these two tasks and show how adequacy and consistency are increased, with virtually no additional cost in terms of model size and training complexity., Comment: Accepted at IWSLT 2022
Published: 2022

9. The focalization domains: An experimental study

Author: Manuella Carnaval, João A. de Moraes, Albert Rilliard, Universidade Federal do Rio de Janeiro (UFRJ), Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), and Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)
Subjects: Entoação, Intonation, Focalization, Brazilian Portuguese, Focalização, Fonologia Prosódica, Português do Brasil, Prosodic Phonology, [SHS.LANGUE]Humanities and Social Sciences/Linguistics
Abstract: RESUMO Este artigo relata um estudo experimental sobre a percepção de domínios focais no Português do Brasil (PB). Utilizando as noções de Foco Amplo e Foco Estreito, propomos um contínuo de extensões abarcadas pela focalização, baseado nos domínios hierárquicos propostos pela teoria da Fonologia Prosódica. São considerados, sobretudo, a palavra fonológica (ω) e o sintagma fonológico (φ), descritos como domínios da entoação no PB. Após a aplicação de um experimento perceptivo multimodal, a análise estatística dos resultados revelou que a palavra fonológica é o domínio preferível de aplicação da focalização no PB, além de apresentar uma interessante sistematicidade nas confusões ocorridas, ao se buscar identificar a incidência do foco, no domínio complexo do sintagma fonológico e de atestar que a adição da modalidade visual não aumenta (nem diminui) significativamente a identificação já eficiente do domínio focal. Além disso, uma análise qualitativa de padrões melódicos selecionados nos permitiu tecer importantes considerações sobre a implementação da focalização na estrutura prosódica do PB. ABSTRACT This paper aims at developing an experimental study about potential restrictions on identifying the length of focused structures in Brazilian Portuguese (BP). Considering Broad Focus and Narrow Focus, we propose a continuum of focalization spans, based on the hierarchical prosodic domains presented in the Prosodic Phonology theory. Specifically, we analyze the intonation domains in BP, the phonological word (ω), and the phonological phrase (φ). After running a multimodal perceptual experiment, statistical results show that narrow elements are best identified than composed ones and that confusions present a systematicity in complex focused structures. As for modality, adding the visual modality does not increase (nor decrease) significantly the already high focus span identification. Furthermore, a qualitative analysis of selected melodic patterns allowed us to develop some relevant remarks about the focalization inscription in the prosodic structure of BP.
Published: 2022

10. Revisiting Multi-Domain Machine Translation

Author: Minh Quang Pham, Josep Maria Crego, François Yvon, Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris-Sud - Paris 11 (UP11)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Université Paris Saclay (COmUE), Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), SYSTRAN, Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), and Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)
Subjects: Neural Machine Translation, Linguistics and Language, Domain adaptation, Machine translation, domain adaptation, Computer science, Communication, 02 engineering and technology, Multi-domain MT, computer.software_genre, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], Computer Science Applications, Large sample, Human-Computer Interaction, Multi domain, Machine Translation, Work (electrical), Artificial Intelligence, Human–computer interaction, 020204 information systems, 0202 electrical engineering, electronic engineering, information engineering, [INFO]Computer Science [cs], 020201 artificial intelligence & image processing, Transfer of learning, computer
Abstract: International audience; When building machine translation systems, one often needs to make the best out of heterogeneous sets of parallel data in training, and to robustly handle inputs from unexpected domains in testing. This multi-domain scenario has attracted a lot of recent work, that fall under the general umbrella of transfer learning. In this study, we revisit multi-domain machine translation, with the aim to formulate the motivations for developing such systems and the associated expectations with respect to performance. Our experiments with a large sample of multi-domain systems show that most of these expectations are hardly met and suggest that further work is needed to better analyze the current behaviour of multi-domain systems and to make them fully hold their promises.
Published: 2021

11. Starting a new treebank? Go SUD! Theoretical and practical benefits of the Surface-Syntactic distributional approach

Author: Gerdes, Kim, Guillaume, Bruno, Kahane, Sylvain, Perrier, Guy, Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Traitement du Langage Parlé (TLP ), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Semantic Analysis of Natural Language (SEMAGRAMME), Inria Nancy - Grand Est, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria)-Department of Natural Language Processing & Knowledge Discovery (LORIA - NLPKD), Laboratoire Lorrain de Recherche en Informatique et ses Applications (LORIA), Centre National de la Recherche Scientifique (CNRS)-Université de Lorraine (UL)-Institut National de Recherche en Informatique et en Automatique (Inria)-Centre National de la Recherche Scientifique (CNRS)-Université de Lorraine (UL)-Institut National de Recherche en Informatique et en Automatique (Inria)-Laboratoire Lorrain de Recherche en Informatique et ses Applications (LORIA), Centre National de la Recherche Scientifique (CNRS)-Université de Lorraine (UL)-Institut National de Recherche en Informatique et en Automatique (Inria)-Centre National de la Recherche Scientifique (CNRS)-Université de Lorraine (UL), Modèles, Dynamiques, Corpus (MoDyCo), Université Paris Nanterre (UPN)-Centre National de la Recherche Scientifique (CNRS), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Institut National de Recherche en Informatique et en Automatique (Inria)-Université de Lorraine (UL)-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-Université de Lorraine (UL)-Centre National de la Recherche Scientifique (CNRS)-Laboratoire Lorrain de Recherche en Informatique et ses Applications (LORIA), Institut National de Recherche en Informatique et en Automatique (Inria)-Université de Lorraine (UL)-Centre National de la Recherche Scientifique (CNRS)-Université de Lorraine (UL)-Centre National de la Recherche Scientifique (CNRS), and Guillaume, Bruno
Subjects: [INFO.INFO-CL] Computer Science [cs]/Computation and Language [cs.CL], [SHS.LANGUE]Humanities and Social Sciences/Linguistics, [SHS.LANGUE] Humanities and Social Sciences/Linguistics, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL]
Abstract: International audience; The paper brings to the fore some advantages to first develop a new treebank in Surface-Syntactic Universal Dependencies (SUD) annotation scheme, even if the goal is to obtain a UD treebank. Theoretical benefits of SUD are presented, as well as UD-compatible SUD innovations. The twoway UD , SUD conversion is explained, as well as the possibility to customize the conversion for a given language. The paper concludes by a practical guide for the development of a SUD treebank.
Published: 2022

12. Speech Technology for Unwritten Languages

Author: Mark Hasegawa-Johnson, Lucas Ondel, Elin Larsen, Shruti Palaskar, Liming Wang, Sebastian Stüker, Francesco Ciannella, Markus Müller, Odette Scharenborg, Rachid Riad, Florian Metze, Pierre Godard, Laurent Besacier, Mingxing Du, Alan W. Black, Danny Merkx, Emmanuel Dupoux, Philip Arthur, Graham Neubig, Delft University of Technology (TU Delft), Groupe d’Étude en Traduction Automatique/Traitement Automatisé des Langues et de la Parole (GETALP), Laboratoire d'Informatique de Grenoble (LIG), Centre National de la Recherche Scientifique (CNRS)-Université Grenoble Alpes (UGA)-Institut polytechnique de Grenoble - Grenoble Institute of Technology (Grenoble INP ), Université Grenoble Alpes (UGA)-Centre National de la Recherche Scientifique (CNRS)-Université Grenoble Alpes (UGA)-Institut polytechnique de Grenoble - Grenoble Institute of Technology (Grenoble INP ), Université Grenoble Alpes (UGA), Carnegie Mellon University [Pittsburgh] (CMU), Department. of Computer Science [Illinois], University of Illinois System, Institute for Anthropomatics (KIT), Karlsruhe Institute of Technology (KIT), Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Brno University of Technology [Brno] (BUT), Johns Hopkins University (JHU), Laboratoire de sciences cognitives et psycholinguistique (LSCP), Département d'Etudes Cognitives - ENS Paris (DEC), École normale supérieure - Paris (ENS-PSL), Université Paris sciences et lettres (PSL)-Université Paris sciences et lettres (PSL)-École normale supérieure - Paris (ENS-PSL), Université Paris sciences et lettres (PSL)-Université Paris sciences et lettres (PSL)-École des hautes études en sciences sociales (EHESS)-Centre National de la Recherche Scientifique (CNRS), Radboud University [Nijmegen], Apprentissage machine et développement cognitif (CoML), Université Paris sciences et lettres (PSL)-Université Paris sciences et lettres (PSL)-École des hautes études en sciences sociales (EHESS)-Centre National de la Recherche Scientifique (CNRS)-Département d'Etudes Cognitives - ENS Paris (DEC), Université Paris sciences et lettres (PSL)-Université Paris sciences et lettres (PSL)-École des hautes études en sciences sociales (EHESS)-Centre National de la Recherche Scientifique (CNRS)-Inria de Paris, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria), The work reported here was started at JSALT 2017 in CMU,Pittsburgh, and was supported by JHU and CMU via grants from Google, Microsoft, Amazon, Facebook, and Apple. This work used the Extreme Science and Engineering Discovery Environment (XSEDE), which is supported by NSF grant number OCI-1053575. Specifically, it used the Bridges system, which is supported by NSF award number ACI-1445606, at the Pittsburgh Supercomputing Center (PSC). OS was partially supported by a Vidi-grant from NWO (276-89-003) and partially by a Delft Technology Fellowship from Delft University of Technology. PG, MM and SS were funded by the French ANR and the German DFG under grant ANR-14-CE35-0002 (BULB project). MD, EL, RR and ED were funded by the European Research Council (ERC-2011-AdG-295810 BOOTPHON), and ANR-10-LABX-0087 IEC and ANR-10-IDEX-0001-02 PSL., ANR-19-P3IA-0001,PRAIRIE,PaRis Artificial Intelligence Research InstitutE(2019), ANR-17-EURE-0017,FrontCog,Frontières en cognition(2017), ANR-10-IDEX-0001,PSL,Paris Sciences et Lettres(2010), ANR-19-P3IA-0003,MIAI,MIAI @ Grenoble Alpes(2019), École normale supérieure - Paris (ENS Paris), Université Paris sciences et lettres (PSL)-Université Paris sciences et lettres (PSL)-École normale supérieure - Paris (ENS Paris), Radboud university [Nijmegen], Inria de Paris, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria)-Laboratoire de sciences cognitives et psycholinguistique (LSCP), Université Pierre Mendès France - Grenoble 2 (UPMF)-Université Joseph Fourier - Grenoble 1 (UJF)-Institut polytechnique de Grenoble - Grenoble Institute of Technology (Grenoble INP)-Institut National Polytechnique de Grenoble (INPG)-Centre National de la Recherche Scientifique (CNRS)-Université Grenoble Alpes (UGA)-Université Pierre Mendès France - Grenoble 2 (UPMF)-Université Joseph Fourier - Grenoble 1 (UJF)-Institut polytechnique de Grenoble - Grenoble Institute of Technology (Grenoble INP)-Institut National Polytechnique de Grenoble (INPG)-Centre National de la Recherche Scientifique (CNRS)-Université Grenoble Alpes (UGA), Université Pierre Mendès France - Grenoble 2 (UPMF)-Université Joseph Fourier - Grenoble 1 (UJF)-Institut polytechnique de Grenoble - Grenoble Institute of Technology (Grenoble INP)-Institut National Polytechnique de Grenoble (INPG)-Centre National de la Recherche Scientifique (CNRS)-Université Grenoble Alpes (UGA), Université Paris-Sud - Paris 11 (UP11)-Université Paris-Saclay-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Centre National de la Recherche Scientifique (CNRS)-Université Paris Saclay (COmUE), Sorbonne Université (SU)-Sorbonne Université (SU)-Centre National de la Recherche Scientifique (CNRS)-Université Paris Saclay (COmUE)-Université Paris-Sud - Paris 11 (UP11)-Université Paris-Saclay-Sorbonne Université - UFR d'Ingénierie (UFR 919), Brno University of Technology, and École normale supérieure - Paris (ENS Paris)-École normale supérieure - Paris (ENS Paris)-École des hautes études en sciences sociales (EHESS)-Centre National de la Recherche Scientifique (CNRS)
Subjects: Acoustics and Ultrasonics, Computer science, Speech synthesis, Semantics, computer.software_genre, unsupervised learning, image retrieval, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], 030507 speech-language pathology & audiology, 03 medical and health sciences, speech synthesis, Computer Science (miscellaneous), Electrical and Electronic Engineering, Everyday life, automatic speech recognition, Speech technology, Speech processing, Linguistics, Language & Communication, Computational Mathematics, Task analysis, Language & Speech Technology, 0305 other medical science, computer, Utterance, Meaning (linguistics)
Abstract: International audience; Speech technology plays an important role in our everyday life. Speech is, among others, used for human-computer interaction, including, for instance, information retrieval and on-line shopping. In the case of an unwritten language, however, speech technology is unfortunately difficult to create, because it cannot be created by the standard combination of pre-trained speech-to-text and text-to-speech subsystems. The research presented in this paper takes the first steps towards speech technology for unwritten languages. Specifically, the aim of this work was 1) to learn speech-to-meaning representations without using text as an intermediate representation, and 2) to test the sufficiency of the learned representations to regenerate speech or translated text, or to retrieve images that depict the meaning of an utterance in an unwritten language. The results suggest that building systems that go directly from speech-to-meaning and from meaning-to-speech, bypassing the need for text, is possible.
Published: 2020

13. A Semi-Automatic Approach to Create Large Gender-and Age-Balanced Speaker Corpora: Usefulness of Speaker Diarization & Identification

Author: Uro, Rémi, Doukhan, David, Rilliard, Albert, Larcher, Laëtitia, Adgharouamane, Anissa-Claire, Tahon, Marie, Laurent, Antoine, Rilliard, Albert, Mesure de l'égalité entre les sexes dans les médias - - GEM2019 - ANR-19-CE38-0012 - AAPG2019 - VALID, Institut National de l'Audiovisuel (INA), Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Expressiveness in Human Centered Data/Media (EXPRESSION), Université de Bretagne Sud (UBS)-MEDIA ET INTERACTIONS (IRISA-D6), Institut de Recherche en Informatique et Systèmes Aléatoires (IRISA), Université de Rennes 1 (UR1), Université de Rennes (UNIV-RENNES)-Université de Rennes (UNIV-RENNES)-Institut National des Sciences Appliquées - Rennes (INSA Rennes), Institut National des Sciences Appliquées (INSA)-Université de Rennes (UNIV-RENNES)-Institut National des Sciences Appliquées (INSA)-Université de Bretagne Sud (UBS)-École normale supérieure - Rennes (ENS Rennes)-Institut National de Recherche en Informatique et en Automatique (Inria)-Télécom Bretagne-CentraleSupélec-Centre National de la Recherche Scientifique (CNRS)-Université de Rennes 1 (UR1), Institut National des Sciences Appliquées (INSA)-Université de Rennes (UNIV-RENNES)-Institut National des Sciences Appliquées (INSA)-Université de Bretagne Sud (UBS)-École normale supérieure - Rennes (ENS Rennes)-Institut National de Recherche en Informatique et en Automatique (Inria)-Télécom Bretagne-CentraleSupélec-Centre National de la Recherche Scientifique (CNRS)-Institut de Recherche en Informatique et Systèmes Aléatoires (IRISA), Institut National des Sciences Appliquées (INSA)-Université de Rennes (UNIV-RENNES)-Institut National des Sciences Appliquées (INSA)-École normale supérieure - Rennes (ENS Rennes)-Institut National de Recherche en Informatique et en Automatique (Inria)-Télécom Bretagne-CentraleSupélec-Centre National de la Recherche Scientifique (CNRS), Laboratoire d'Informatique de l'Université du Mans (LIUM), Le Mans Université (UM), Spécinov SARL, and ANR-19-CE38-0012,GEM,Mesure de l'égalité entre les sexes dans les médias(2019)
Subjects: semi-automatic processing, speaker identification, age-balanced, diachrony, corpus creation, speaker corpus, [INFO]Computer Science [cs], [SHS.LANGUE]Humanities and Social Sciences/Linguistics, [INFO] Computer Science [cs], [SHS.LANGUE] Humanities and Social Sciences/Linguistics, gender-balanced, diarization
Abstract: International audience; This paper presents a semi-automatic approach to create a diachronic corpus of voices balanced for speaker's age, gender, and recording period, according to 32 categories (2 genders, 4 age ranges and 4 recording periods). Corpora were selected at French National Institute of Audiovisual (INA) to obtain at least 30 speakers per category (a total of 960 speakers; only 874 have be found yet). For each speaker, speech excerpts were extracted from audiovisual documents using an automatic pipeline consisting of speech detection, background music and overlapped speech removal and speaker diarization, used to present clean speaker segments to human annotators identifying target speakers. This pipeline proved highly effective, cutting down manual processing by a factor of ten. Evaluation of the quality of the automatic processing and of the final output is provided. It shows the automatic processing compare to up-to-date process, and that the output provides high quality speech for most of the selected excerpts. This method shows promise for creating large corpora of known target speakers.
Published: 2022

14. Focus types in Brazilian Portuguese: Multimodal production and perception

Author: Manuella Carnaval, João Antônio de Moraes, Albert Rilliard, Universidade Federal do Rio de Janeiro (UFRJ), Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), and Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)
Subjects: Linguistics and Language, focalization, Brazilian Portuguese, [SHS.LANGUE]Humanities and Social Sciences/Linguistics, audiovisual prosody
Abstract: International audience; This paper aims at describing prosodic focalization as a multimodal phenomenon in Brazilian Portuguese, evaluating the role of two modalities in focus production and perception: audio (A), visual (V), and their combined audiovisual presentation (AV). Five focus types are considered, according to their semantic-pragmatic values: (a) in declarative sentences: (i) IF - informational focus (answer to a previous question, conveying new information), (ii) CF- contrastive (strong) focus (correction of information considered wrong); (iii) ATF - attenuated (weak) focus (proposition of an alternative solution to previous information); (b) in interrogative sentences: (i) INTF - interrogative focus (a new information is requested in the question); (ii) SF - surprise focus (one casts doubt on a previous information). Also, structural factors were evaluated, as focus extension and position in the sentence. After running a multimodal perceptual experiment and developing an acoustic and visual analysis on focus production, results show that multimodality plays a relevant role in focus production and perception. Different acoustic and visual parameters, or configuration of parameters, contribute to conveying distinct meanings, according to each focus type.; Este artigo descreve a focalização prosódica como um fenômeno multimodal no português do Brasil, investigando a relevância de duas modalidades na produção e percepção do foco: auditiva (A), visual (V), além de sua apresentação audiovisual (AV). Cinco tipos de foco são considerados, de acordo com seus valores semântico-pragmáticos: (a) em enunciados assertivos: (i) FI - foco informacional (uma resposta a uma pergunta prévia, constituindo informação nova), (ii) FC - foco contrastivo (correção de uma informação errada), (iii) FAT - foco atenuado (proposição de uma solução alternativa para uma afirmação prévia); (b) em enunciados interrogativos: (i) FINT - foco interrogativo (uma nova informação requisitada em uma pergunta); (ii) FE - foco com estranheza (uma informação prévia é colocada em dúvida). Também foram investigadas a extensão e a posição do foco na sentença. Após a aplicação de um experimento perceptivo e de uma análise multimodal, os resultados mostraram que a multimodalidade apresenta relevância na produção e percepção do foco. No entanto, diferentes parâmetros acústicos e visuais, ou a sua combinação, contribuem para a transmissão de diferentes significados, segundo cada tipo de foco.
Published: 2022

15. BLOOM: A 176B-Parameter Open-Access Multilingual Language Model

Author: Workshop, BigScience, Scao, Teven Le, Fan, Angela, Akiki, Christopher, Pavlick, Ellie, Ilić, Suzana, Hesslow, Daniel, Castagné, Roman, Luccioni, Alexandra Sasha, Yvon, François, Gallé, Matthias, Tow, Jonathan, Rush, Alexander M., Biderman, Stella, Webson, Albert, Ammanamanchi, Pawan Sasanka, Wang, Thomas, Sagot, Benoît, Muennighoff, Niklas, del Moral, Albert Villanova, Ruwase, Olatunji, Bawden, Rachel, Bekman, Stas, McMillan-Major, Angelina, Beltagy, Iz, Nguyen, Huu, Saulnier, Lucile, Tan, Samson, Suarez, Pedro Ortiz, Sanh, Victor, Laurençon, Hugo, Jernite, Yacine, Launay, Julien, Mitchell, Margaret, Raffel, Colin, Gokaslan, Aaron, Simhi, Adi, Soroa, Aitor, Aji, Alham Fikri, Alfassy, Amit, Rogers, Anna, Nitzav, Ariel Kreisberg, Xu, Canwen, Mou, Chenghao, Emezue, Chris, Klamm, Christopher, Leong, Colin, van Strien, Daniel, Adelani, David Ifeoluwa, Radev, Dragomir, Ponferrada, Eduardo González, Levkovizh, Efrat, Kim, Ethan, Natan, Eyal Bar, De Toni, Francesco, Dupont, Gérard, Kruszewski, Germán, Pistilli, Giada, Elsahar, Hady, Benyamina, Hamza, Tran, Hieu, Yu, Ian, Abdulmumin, Idris, Johnson, Isaac, Gonzalez-Dios, Itziar, de la Rosa, Javier, Chim, Jenny, Dodge, Jesse, Zhu, Jian, Chang, Jonathan, Frohberg, Jörg, Tobing, Joseph, Bhattacharjee, Joydeep, Almubarak, Khalid, Chen, Kimbo, Lo, Kyle, Von Werra, Leandro, Weber, Leon, Phan, Long, allal, Loubna Ben, Tanguy, Ludovic, Dey, Manan, Muñoz, Manuel Romero, Masoud, Maraim, Grandury, María, Šaško, Mario, Huang, Max, Coavoux, Maximin, Singh, Mayank, Jiang, Mike Tian-Jian, Vu, Minh Chien, Jauhar, Mohammad A., Ghaleb, Mustafa, Subramani, Nishant, Kassner, Nora, Khamis, Nurulaqilla, Nguyen, Olivier, Espejel, Omar, de Gibert, Ona, Villegas, Paulo, Henderson, Peter, Colombo, Pierre, Amuok, Priscilla, Lhoest, Quentin, Harliman, Rheza, Bommasani, Rishi, López, Roberto Luis, Ribeiro, Rui, Osei, Salomey, Pyysalo, Sampo, Nagel, Sebastian, Bose, Shamik, Muhammad, Shamsuddeen Hassan, Sharma, Shanya, Longpre, Shayne, Nikpoor, Somaieh, Silberberg, Stanislav, Pai, Suhas, Zink, Sydney, Torrent, Tiago Timponi, Schick, Timo, Thrush, Tristan, Danchev, Valentin, Nikoulina, Vassilina, Laippala, Veronika, Lepercq, Violette, Prabhu, Vrinda, Alyafeai, Zaid, Talat, Zeerak, Raja, Arun, Heinzerling, Benjamin, Si, Chenglei, Taşar, Davut Emre, Salesky, Elizabeth, Mielke, Sabrina J., Lee, Wilson Y., Sharma, Abheesht, Santilli, Andrea, Chaffin, Antoine, Stiegler, Arnaud, Datta, Debajyoti, Szczechla, Eliza, Chhablani, Gunjan, Wang, Han, Pandey, Harshit, Strobelt, Hendrik, Fries, Jason Alan, Rozen, Jos, Gao, Leo, Sutawika, Lintang, Bari, M Saiful, Al-shaibani, Maged S., Manica, Matteo, Nayak, Nihal, Teehan, Ryan, Albanie, Samuel, Shen, Sheng, Ben-David, Srulik, Bach, Stephen H., Kim, Taewoon, Bers, Tali, Fevry, Thibault, Neeraj, Trishala, Thakker, Urmish, Raunak, Vikas, Tang, Xiangru, Yong, Zheng-Xin, Sun, Zhiqing, Brody, Shaked, Uri, Yallow, Tojarieh, Hadar, Roberts, Adam, Chung, Hyung Won, Tae, Jaesung, Phang, Jason, Press, Ofir, Li, Conglong, Narayanan, Deepak, Bourfoune, Hatim, Casper, Jared, Rasley, Jeff, Ryabinin, Max, Mishra, Mayank, Zhang, Minjia, Shoeybi, Mohammad, Peyrounette, Myriam, Patry, Nicolas, Tazi, Nouamane, Sanseviero, Omar, von Platen, Patrick, Cornette, Pierre, Lavallée, Pierre François, Lacroix, Rémi, Rajbhandari, Samyam, Gandhi, Sanchit, Smith, Shaden, Requena, Stéphane, Patil, Suraj, Dettmers, Tim, Baruwa, Ahmed, Singh, Amanpreet, Cheveleva, Anastasia, Ligozat, Anne-Laure, Subramonian, Arjun, Névéol, Aurélie, Lovering, Charles, Garrette, Dan, Tunuguntla, Deepak, Reiter, Ehud, Taktasheva, Ekaterina, Voloshina, Ekaterina, Bogdanov, Eli, Winata, Genta Indra, Schoelkopf, Hailey, Kalo, Jan-Christoph, Novikova, Jekaterina, Forde, Jessica Zosa, Clive, Jordan, Kasai, Jungo, Kawamura, Ken, Hazan, Liam, Carpuat, Marine, Clinciu, Miruna, Kim, Najoung, Cheng, Newton, Serikov, Oleg, Antverg, Omer, van der Wal, Oskar, Zhang, Rui, Zhang, Ruochen, Gehrmann, Sebastian, Mirkin, Shachar, Pais, Shani, Shavrina, Tatiana, Scialom, Thomas, Yun, Tian, Limisiewicz, Tomasz, Rieser, Verena, Protasov, Vitaly, Mikhailov, Vladislav, Pruksachatkun, Yada, Belinkov, Yonatan, Bamberger, Zachary, Kasner, Zdeněk, Rueda, Alice, Pestana, Amanda, Feizpour, Amir, Khan, Ammar, Faranak, Amy, Santos, Ana, Hevia, Anthony, Unldreaj, Antigona, Aghagol, Arash, Abdollahi, Arezoo, Tammour, Aycha, HajiHosseini, Azadeh, Behroozi, Bahareh, Ajibade, Benjamin, Saxena, Bharat, Ferrandis, Carlos Muñoz, McDuff, Daniel, Contractor, Danish, Lansky, David, David, Davis, Kiela, Douwe, Nguyen, Duong A., Tan, Edward, Baylor, Emi, Ozoani, Ezinwanne, Mirza, Fatima, Ononiwu, Frankline, Rezanejad, Habib, Jones, Hessie, Bhattacharya, Indrani, Solaiman, Irene, Sedenko, Irina, Nejadgholi, Isar, Passmore, Jesse, Seltzer, Josh, Sanz, Julio Bonis, Dutra, Livia, Samagaio, Mairon, Elbadri, Maraim, Mieskes, Margot, Gerchick, Marissa, Akinlolu, Martha, McKenna, Michael, Qiu, Mike, Ghauri, Muhammed, Burynok, Mykola, Abrar, Nafis, Rajani, Nazneen, Elkott, Nour, Fahmy, Nour, Samuel, Olanrewaju, An, Ran, Kromann, Rasmus, Hao, Ryan, Alizadeh, Samira, Shubber, Sarmad, Wang, Silas, Roy, Sourav, Viguier, Sylvain, Le, Thanh, Oyebade, Tobi, Le, Trieu, Yang, Yoyo, Nguyen, Zach, Kashyap, Abhinav Ramesh, Palasciano, Alfredo, Callahan, Alison, Shukla, Anima, Miranda-Escalada, Antonio, Singh, Ayush, Beilharz, Benjamin, Wang, Bo, Brito, Caio, Zhou, Chenxi, Jain, Chirag, Xu, Chuxin, Fourrier, Clémentine, Periñán, Daniel León, Molano, Daniel, Yu, Dian, Manjavacas, Enrique, Barth, Fabio, Fuhrimann, Florian, Altay, Gabriel, Bayrak, Giyaseddin, Burns, Gully, Vrabec, Helena U., Bello, Imane, Dash, Ishani, Kang, Jihyun, Giorgi, John, Golde, Jonas, Posada, Jose David, Sivaraman, Karthik Rangasai, Bulchandani, Lokesh, Liu, Lu, Shinzato, Luisa, de Bykhovetz, Madeleine Hahn, Takeuchi, Maiko, Pàmies, Marc, Castillo, Maria A, Nezhurina, Marianna, Sänger, Mario, Samwald, Matthias, Cullan, Michael, Weinberg, Michael, De Wolf, Michiel, Mihaljcic, Mina, Liu, Minna, Freidank, Moritz, Kang, Myungsun, Seelam, Natasha, Dahlberg, Nathan, Broad, Nicholas Michio, Muellner, Nikolaus, Fung, Pascale, Haller, Patrick, Chandrasekhar, Ramya, Eisenberg, Renata, Martin, Robert, Canalli, Rodrigo, Su, Rosaline, Su, Ruisi, Cahyawijaya, Samuel, Garda, Samuele, Deshmukh, Shlok S, Mishra, Shubhanshu, Kiblawi, Sid, Ott, Simon, Sang-aroonsiri, Sinee, Kumar, Srishti, Schweter, Stefan, Bharati, Sushil, Laud, Tanmay, Gigant, Théo, Kainuma, Tomoya, Kusa, Wojciech, Labrak, Yanis, Bajaj, Yash Shailesh, Venkatraman, Yash, Xu, Yifan, Xu, Yingxin, Xu, Yu, Tan, Zhe, Xie, Zhongli, Ye, Zifan, Bras, Mathilde, Belkada, Younes, Wolf, Thomas, Natural Language Processing : representations, inference and semantics (SYNALP), Department of Natural Language Processing & Knowledge Discovery (LORIA - NLPKD), Laboratoire Lorrain de Recherche en Informatique et ses Applications (LORIA), Institut National de Recherche en Informatique et en Automatique (Inria)-Université de Lorraine (UL)-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-Université de Lorraine (UL)-Centre National de la Recherche Scientifique (CNRS)-Laboratoire Lorrain de Recherche en Informatique et ses Applications (LORIA), Institut National de Recherche en Informatique et en Automatique (Inria)-Université de Lorraine (UL)-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-Université de Lorraine (UL)-Centre National de la Recherche Scientifique (CNRS), Automatic Language Modelling and ANAlysis & Computational Humanities (ALMAnaCH), Inria de Paris, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Traitement du Langage Parlé (TLP ), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Data and Web Science Group [Mannheim Univ] (DWS), Universität Mannheim, Information, Langue Ecrite et Signée (ILES), Avignon Université (AU), Laboratoire Informatique d'Avignon (LIA), Avignon Université (AU)-Centre d'Enseignement et de Recherche en Informatique - CERI, and Zenidoc
Subjects: FOS: Computer and information sciences, Computer Science - Computation and Language, Computation and Language (cs.CL), [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], [INFO.INFO-AI]Computer Science [cs]/Artificial Intelligence [cs.AI]
Abstract: Large language models (LLMs) have been shown to be able to perform new tasks based on a few demonstrations or natural language instructions. While these capabilities have led to widespread adoption, most LLMs are developed by resource-rich organizations and are frequently kept from the public. As a step towards democratizing this powerful technology, we present BLOOM, a 176B-parameter open-access language model designed and built thanks to a collaboration of hundreds of researchers. BLOOM is a decoder-only Transformer language model that was trained on the ROOTS corpus, a dataset comprising hundreds of sources in 46 natural and 13 programming languages (59 in total). We find that BLOOM achieves competitive performance on a wide variety of benchmarks, with stronger results after undergoing multitask prompted finetuning. To facilitate future research and applications using LLMs, we publicly release our models and code under the Responsible AI License.
Published: 2022
Full Text: View/download PDF

16. Evaluating Subtitle Segmentation for End-to-end Generation Systems

Author: Alina Karakanta, François Buet, Mauro Cettolo, François Yvon, Fondazione Bruno Kessler [Trento, Italy] (FBK), Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Européen, ELDA, and European Project: 952026,EC:H2020 | RIA,HumanE AI Network(2020)
Subjects: [INFO.INFO-TT]Computer Science [cs]/Document and Text Processing, FOS: Computer and information sciences, Segmentation, Computer Science - Computation and Language, Subtitling, Metric, Evaluation, Computation and Language (cs.CL)
Abstract: Subtitles appear on screen as short pieces of text, segmented based on formal constraints (length) and syntactic/semantic criteria. Subtitle segmentation can be evaluated with sequence segmentation metrics against a human reference. However, standard segmentation metrics cannot be applied when systems generate outputs different than the reference, e.g. with end-to-end subtitling systems. In this paper, we study ways to conduct reference-based evaluations of segmentation accuracy irrespective of the textual content. We first conduct a systematic analysis of existing metrics for evaluating subtitle segmentation. We then introduce $Sigma$, a new Subtitle Segmentation Score derived from an approximate upper-bound of BLEU on segmentation boundaries, which allows us to disentangle the effect of good segmentation from text quality. To compare $Sigma$ with existing metrics, we further propose a boundary projection method from imperfect hypotheses to the true reference. Results show that all metrics are able to reward high quality output but for similar outputs system ranking depends on each metric's sensitivity to error type. Our thorough analyses suggest $Sigma$ is a promising segmentation candidate but its reliability over other segmentation metrics remains to be validated through correlations with human judgements., Comment: Accepted at LREC 2022
Published: 2022
Full Text: View/download PDF

17. Effects of F0 movements, intensity, and duration in the perceptual identification of Brazilian Portuguese wh-questions and wh-exclamations

Author: Luma da Silva Miranda, João Antônio de Moraes, Albert Rilliard, Universidade Federal do Rio de Janeiro (UFRJ), Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), and Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)
Subjects: intonation, Linguistics and Language, Brazilian Portuguese, [SHS.LANGUE]Humanities and Social Sciences/Linguistics, perception, wh-question, wh-exclamation
Abstract: International audience; ABSTRACT This study presents a perceptual analysis of the Brazilian Portuguese wh-question and wh-exclamation intonational contours to discriminate their acoustic and perceptual features. The corpus of this study is composed of the sentence “Como você sabe” (“How do you know” vs. “How clever you are!”), which was produced with both speech acts. Two perceptual identification experiments were designed to assess the subjects’ ability to identify these speech acts based on their prosodic characteristics, as well as the perceptual relevance of specific prosodic cues in the recognition of wh-questions and wh-exclamations. The results of the first perceptual test indicated that Brazilian listeners can identify these two speech acts by intonation only, whereas the second test showed that F0, duration and intensity cues contribute to the perceptual identification of the speech acts. Stimuli with a falling F0 movement in the last stressed syllable tend to be interpreted as wh-questions, whereas stimuli with a slightly rising F0 movement tend to be judged as wh-exclamations.; RESUMO Este estudo apresenta uma análise perceptiva da entoação da questão parcial e da exclamação do português brasileiro com o objetivo de discriminar suas características acústicas e perceptivas. O corpus deste estudo é constituído pela frase “Como você sabe”, que foi produzida com os dois atos de fala. Aplicaram-se dois testes de percepção para avaliar a capacidade dos sujeitos em identificar esses atos de fala baseados nas suas características prosódicas, bem como a relevância perceptiva de pistas prosódicas específicas que atuam no reconhecimento da questão parcial e da exclamação. Os resultados do primeiro teste perceptivo aplicado com vinte e cinco ouvintes indicaram que eles identificam esses dois atos de fala somente pela entoação, enquanto os do segundo teste aplicado com vinte participantes mostraram que a configuração da F0 no núcleo das sentenças e as informações de intensidade e duração contribuem para o reconhecimento perceptivo desses atos de fala. Estímulos com movimento de F0 descendente na última sílaba tônica tendem a ser interpretados como questão parcial, enquanto os estímulos com movimento de F0 ascendente de baixa amplitude tendem a ser julgados como exclamação.
Published: 2022

18. A corpus-based study of the distribution of word-final schwa in Standard French and what it teaches us about its phonological status

Author: Mathilde Hutin, Adèle Jatteau, Ioana Vasilescu, Lori Lamel, Yaru Wu, Martine Adda-Decker, Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Traitement du Langage Parlé (TLP ), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Savoirs, Textes, Langage (STL) - UMR 8163 (STL), Université de Lille-Centre National de la Recherche Scientifique (CNRS), LPP - Laboratoire de Phonétique et Phonologie - UMR 7018 (LPP), Université Sorbonne Nouvelle - Paris 3-Centre National de la Recherche Scientifique (CNRS), Université de Caen Normandie - UFR Humanités et Sciences Sociales (UNICAEN UFR HSS), Université de Caen Normandie (UNICAEN), Normandie Université (NU)-Normandie Université (NU), Laboratoire Interdisciplinaire des Sciences du Numérique [LISN], Savoirs, Textes, Langage (STL) - UMR 8163 [STL], LPP - Laboratoire de Phonétique et Phonologie - UMR 7018 [LPP], CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), and CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)
Subjects: Standard French, Standard french, Pronunciation variation, automatic alignment, schwa, P1-1091, large corpora, [SCCO.LING]Cognitive science/Linguistics, epenthesis, Large corpora, Automatic alignment, pronunciation variation, PC1-5498, Schwa, [SHS.LANGUE]Humanities and Social Sciences/Linguistics, Philology. Linguistics, Epenthesis, ComputingMilieux_MISCELLANEOUS, Romanic languages
Abstract: What is commonly considered as an epenthetic vowel can actually refer to at least two different realities: phonological epenthesis or phonetic excrescence. French schwa, noted [ә], is a vowel alternating with zero and limited to unstressed syllables that can appear word-internally or word-finally. This paper presents an extensive description of the distribution of word-final schwa in Standard French in order to shed light on its nature: is it an intrusive vowel or a full epenthetic vowel? To that extent, three large corpora of French containing more than 110 hours of speech were used to establish the presence of word-final schwa as a function of sociolinguistics, orthography, phonotactics and phonetics. Our conclusions are that word-final schwa is impacted by speech style, gender, orthography, phonotactics (i.e., the number of adjacent consonants and their sonority profile), and the phonological properties of the codas. However, speech rate does not impact word-final schwa realization. The specific results lead us to suggest that word-final schwa in Standard French shares similarities with intrusive vowels but ultimately behaves like a legit epenthetic vowel.
Published: 2021

19. Screening Gender Transfer in Neural Machine Translation

Author: Guillaume Wisniewski, Lichao Zhu, Nicolas Ballier, François Yvon, Laboratoire de Linguistique Formelle (LLF - UMR7110), Centre National de la Recherche Scientifique (CNRS)-Université Paris Cité (UPCité), Centre de Linguistique Inter-langues, de Lexicologie, de Linguistique Anglaise et de Corpus (CLILLAC-ARP (URP_3967)), Université Paris Cité (UPCité), Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Association for computational linguistics, Laboratoire de Linguistique Formelle (LLF UMR7110), Centre National de la Recherche Scientifique (CNRS)-Université de Paris (UP), Université de Paris (UP), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), This work was partially funded by the NeuroViz project subsidized by the Ile-de-France Region, and supported in part by the 2020 émergence research project SPECTRANS., and EMNLP
Subjects: FOS: Computer and information sciences, [INFO.INFO-TT]Computer Science [cs]/Document and Text Processing, Neural Machine Translation, Computer Science - Computation and Language, Probing, Gender bias, Transformer architectures, [INFO]Computer Science [cs], [SHS.LANGUE]Humanities and Social Sciences/Linguistics, Computation and Language (cs.CL)
Abstract: This paper aims at identifying the information flow in state-of-the-art machine translation systems, taking as example the transfer of gender when translating from French into English. Using a controlled set of examples, we experiment several ways to investigate how gender information circulates in a encoder-decoder architecture considering both probing techniques as well as interventions on the internal representations used in the MT system. Our results show that gender information can be found in all token representations built by the encoder and the decoder and lead us to conclude that there are multiple pathways for gender transfer., Accepted at BlackBoxNLP'2021
Published: 2021

20. One Source, Two Targets: Challenges and Rewards of Dual Decoding

Author: Xu, Jitao, Yvon, François, Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Bourse régionale IDF, CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), and CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)
Subjects: FOS: Computer and information sciences, MachineTranslation, Decoding Algorithms, Computer Science - Computation and Language, 0202 electrical engineering, electronic engineering, information engineering, 020201 artificial intelligence & image processing, [INFO]Computer Science [cs], 02 engineering and technology, 010501 environmental sciences, 01 natural sciences, Computation and Language (cs.CL), [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], 0105 earth and related environmental sciences
Abstract: Machine translation is generally understood as generating one target text from an input source document. In this paper, we consider a stronger requirement: to jointly generate two texts so that each output side effectively depends on the other. As we discuss, such a device serves several practical purposes, from multi-target machine translation to the generation of controlled variations of the target text. We present an analysis of possible implementations of dual decoding, and experiment with four applications. Viewing the problem from multiple angles allows us to better highlight the challenges of dual decoding and to also thoroughly analyze the benefits of generating matched, rather than independent, translations., Comment: Accepted at EMNLP 2021
Published: 2021

21. Graph Algorithms for Multiparallel Word Alignment

Author: Imani, Ayyoob, Sabet, Masoud Jalili, ��enel, L��tfi Kerem, Dufter, Philipp, Yvon, Fran��ois, Sch��tze, Hinrich, Centre for Information and Language Processing (CIS ), Ludwigs-Maximilians-Universität Munich, Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Association for Computational Linguistics, Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), and Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)
Subjects: FOS: Computer and information sciences, Computer Science - Computation and Language, multilingual natural language processing, graph neural networks, 02 engineering and technology, 010501 environmental sciences, 01 natural sciences, [INFO.INFO-TT]Computer Science [cs]/Document and Text Processing, traitement automatique des langues multilingue, 0202 electrical engineering, electronic engineering, information engineering, 020201 artificial intelligence & image processing, word alignments, Computation and Language (cs.CL), alignement de mots, 0105 earth and related environmental sciences
Abstract: With the advent of end-to-end deep learning approaches in machine translation, interest in word alignments initially decreased; however, they have again become a focus of research more recently. Alignments are useful for typological research, transferring formatting like markup to translated texts, and can be used in the decoding of machine translation systems. At the same time, massively multilingual processing is becoming an important NLP scenario, and pretrained language and machine translation models that are truly multilingual are proposed. However, most alignment algorithms rely on bitexts only and do not leverage the fact that many parallel corpora are multiparallel. In this work, we exploit the multiparallelity of corpora by representing an initial set of bilingual alignments as a graph and then predicting additional edges in the graph. We present two graph algorithms for edge prediction: one inspired by recommender systems and one based on network link prediction. Our experimental results show absolute improvements in $F_1$ of up to 28% over the baseline bilingual word aligner in different datasets., EMNLP 2021
Published: 2021

22. Toward Genre Adapted Closed Captioning

Author: François Buet, François Yvon, Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), BPI/Rosetta, CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), and CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)
Subjects: Closed captioning, Machine translation, Text simplification, Computer science, business.industry, Contrast (statistics), Automatic Captioning, 020206 networking & telecommunications, 02 engineering and technology, computer.software_genre, Speech Transcription, [INFO.INFO-TT]Computer Science [cs]/Document and Text Processing, 030507 speech-language pathology & audiology, 03 medical and health sciences, Text Simplification, 0202 electrical engineering, electronic engineering, information engineering, Automatic speech, Speech transcription, Artificial intelligence, 0305 other medical science, Set (psychology), Adaptation (computer science), business, computer, Natural language processing
Abstract: International audience; This paper studies the generation of intralingual closed captions from automatic speech transcripts, with the aim to assess techniques for multi-genre captioning. Captions and subtitles greatly vary in form and content depending on the programs genres and subtitling styles, resulting for instance in significantly different compression rates and lexical content. Borrowing ideas from the multi-domain machine translation literature, we implement and contrast several adaptation methods on a diverse set of programs broadcast on the French public TV. Our results show that such multi-domain adaption techniques are effective and help to improve our automatic subtitling system.; Cet article étudie la génération de sous-titres monolingues à partir de transcriptions vocales automatiques, dans le but d'évaluer les techniques de sous-titrage multi-genres. La forme et le contenu des sous-titres varient considérablement en fonction des genres de programmes et des styles de sous-titrage, ce qui se traduit par exemple par des taux de compression et des contenus lexicaux très différents. En empruntant des idées à la littérature de la traduction automatique multi-domaine, nous mettons en œuvre et comparons plusieurs méthodes d'adaptation sur un ensemble diversifié de programmes diffusés sur la télévision publique française. Nos résultats montrent que ces techniques d'adaptation multi-domaines sont efficaces et permettent d'améliorer notre système de sous-titrage automatique.
Published: 2021

23. Prosodic Boundary Prediction Model for Vietnamese Text-To-Speech

Author: Trang Nguyen, Nguyen Hoang Ky, Christophe d'Alessandro, Nguyen Thi Thu Trang, Albert Rilliard, Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Sciences et Technologies des Langues (STL), Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec, Lutheries - Acoustique - Musique (IJLRDA-LAM), Institut Jean le Rond d'Alembert (DALEMBERT), Université Pierre et Marie Curie - Paris 6 (UPMC)-Centre National de la Recherche Scientifique (CNRS)-Université Pierre et Marie Curie - Paris 6 (UPMC)-Centre National de la Recherche Scientifique (CNRS), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Institut Jean Le Rond d'Alembert (DALEMBERT), Sorbonne Université (SU)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université (SU)-Centre National de la Recherche Scientifique (CNRS), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), and Centre National de la Recherche Scientifique (CNRS)-Université Pierre et Marie Curie - Paris 6 (UPMC)-Centre National de la Recherche Scientifique (CNRS)-Université Pierre et Marie Curie - Paris 6 (UPMC)
Subjects: Computer science, Vietnamese, Speech recognition, pause prediction, Prosody modeling, Boundary (topology), 020206 networking & telecommunications, Speech synthesis, 02 engineering and technology, computer.software_genre, language.human_language, 030507 speech-language pathology & audiology, 03 medical and health sciences, speech synthesis, [INFO.INFO-TS]Computer Science [cs]/Signal and Image Processing, [INFO.INFO-SD]Computer Science [cs]/Sound [cs.SD], 0202 electrical engineering, electronic engineering, information engineering, language, prosodic boundary, Text-To-Speech, [INFO.INFO-HC]Computer Science [cs]/Human-Computer Interaction [cs.HC], [SHS.LANGUE]Humanities and Social Sciences/Linguistics, 0305 other medical science, computer
Abstract: International audience; This research aims to build a prosodic boundary prediction model for improving the naturalness of Vietnamese speech synthesis. This model can be used directly to predict prosodic boundaries in the synthesis phase of the statistical parametric or end-to-end speech systems. Beside conventional features related to Part-Of-Speech (POS), this paper proposes two efficient features to predict prosodic boundaries: syntactic blocks and syntactic links, based on a thorough analysis of a Vietnamese dataset. Syntactic blocks are syntactic phrases whose sizes are bounded in their constituent syntactic tree. A syntactic link of two adjacent words is calculated based on the distance between them in the syntax tree. The experimental results show that the two proposed predictors improve the quality of the boundary prediction model using a decision tree classification algorithm, about 36.4% (F1 score) higher than the model with only POS features. The final boundary prediction model with POS, syntactic block, and syntactic link features using the LightGBM algorithm gives the best F1-score results at 87.0% in test data. The proposed model helps the TTS systems, developed by either HMM-based, DNN-based, or End-to-end speech synthesis techniques, improve about 0.3 MOS points (i.e. 6 to 10%) compared to the ones without the proposed model.
Published: 2021

24. Operation LiLi: Using Crowd-Sourced Data and Automatic Alignment to Investigate the Phonetics and Phonology of Less-Resourced Languages

Author: Mathilde Hutin, Marc Tang, Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Traitement du Langage Parlé (TLP ), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Éco-Anthropologie (EA), Muséum national d'Histoire naturelle (MNHN)-Centre National de la Recherche Scientifique (CNRS), and ANR-20-CE27-0021,EVOGRAM,L'effet des facteurs linguistiques et non-linguistiques sur l'évolution des systèmes de classification nominale(2020)
Subjects: Linguistics and Language, [SCCO.LING]Cognitive science/Linguistics, vowel dispersion theory, Lingua Libre, phonetics, crowd-sourced data, Language and Linguistics
Abstract: International audience; Less-resourced languages are usually left out of phonetic studies based on large corpora. We contribute to the recent efforts to fill this gap by assessing how to use open-access, crowd-sourced audio data from Lingua Libre for phonetic research. Lingua Libre is a participative linguistic library developed by Wikimedia France in 2015. It contains more than 670k recordings in approximately 150 languages across nearly 740 speakers. As a proof of concept, we consider the Inventory Size Hypothesis, which predicts that, in a given system, variation in the realization of each vowel will be inversely related to the number of vowel categories. We investigate data from 10 languages with various numbers of vowel categories, i.e., German, Afrikaans, French, Catalan, Italian, Romanian, Polish, Russian, Spanish, and Basque. Audio files are extracted from Lingua Libre to be aligned and segmented using the Munich Automatic Segmentation System. Information on the formants of the vowel segments is then extracted to measure how vowels expand in the acoustic space and whether this is correlated with the number of vowel categories in the language. The results provide valuable insight into the question of vowel dispersion and demonstrate the wealth of information that crowd-sourced data has to offer.
Published: 2022

25. Coherent structure identification in turbulent channel flow using latent Dirichlet allocation

Author: Mohamed Frihat, Yann Fraigneau, Lionel Mathelin, Bérengère Podvin, François Yvon, Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Traitement du Langage Parlé (TLP ), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), DAta science, TrAnsition, Fluid instabiLity, contrOl, Turbulence (DATAFLOT), Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Université Paris-Sud - Paris 11 (UP11)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Université Paris Saclay (COmUE), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Accompagnement et Soutien aux Activités de Recherche & Développement (ASARD), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), and Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL)
Subjects: Scale (ratio), Turbulence, Mechanical Engineering, Fluid Dynamics (physics.flu-dyn), FOS: Physical sciences, Statistical model, Physics - Fluid Dynamics, Latent variable, Condensed Matter Physics, 01 natural sciences, Latent Dirichlet allocation, 010305 fluids & plasmas, Physics::Fluid Dynamics, Set (abstract data type), [SPI]Engineering Sciences [physics], symbols.namesake, Flow (mathematics), Mechanics of Materials, 0103 physical sciences, symbols, 010306 general physics, Cluster analysis, Algorithm, ComputingMilieux_MISCELLANEOUS
Abstract: Identification of coherent structures is an essential step to describe and model turbulence generation mechanisms in wall-bounded flows. To this end, we present a clustering method based on latent Dirichlet allocation (LDA), a generative probabilistic model for collection of discrete data. The method is applied to a set of snapshots featuring the Reynolds stress (. Both two- and three-dimensional analysis show that LDA provides a robust and compact flow description in terms of a combination of motifs, which are latent variables inferred from the set of snapshots. We find that the characteristics of the motifs scale with the wall distance, in agreement with the wall-attached eddy hypothesis of Townsend (Physics of Fluids, 1961, pp. 97–120). Latent Dirichlet allocation motifs can be used to reconstruct fields with an efficiency that can be compared with the proper orthogonal decomposition (POD). Moreover, the LDA model makes it possible to generate a collection of synthetic fields that captures the intermittent characteristics of the original dataset more clearly than its POD-generated counterpart. These findings highlight the potential of LDA for turbulent flow analysis, efficient reconstruction of actual fields and production of a new field with suitable statistics.
Published: 2021

26. Can You Traducir This? Machine Translation for Code-Switched Input

Author: Jitao Xu, François Yvon, Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Projet Régional IDF 'Tremplin PhD', Association for Computational Linguistics, Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), and Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)
Subjects: FOS: Computer and information sciences, Neural Machine Translation, Machine translation, Computer science, 02 engineering and technology, 010501 environmental sciences, computer.software_genre, 01 natural sciences, Code (semiotics), [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], Task (project management), Writing Aid, Phenomenon, 0202 electrical engineering, electronic engineering, information engineering, 0105 earth and related environmental sciences, Code Switching, Focus (computing), Training set, Computer Science - Computation and Language, business.industry, Code-switching, 020201 artificial intelligence & image processing, Artificial intelligence, business, computer, Computation and Language (cs.CL), Natural language processing
Abstract: Code-Switching (CSW) is a common phenomenon that occurs in multilingual geographic or social contexts, which raises challenging problems for natural language processing tools. We focus here on Machine Translation (MT) of CSW texts, where we aim to simultaneously disentangle and translate the two mixed languages. Due to the lack of actual translated CSW data, we generate artificial training data from regular parallel texts. Experiments show this training strategy yields MT systems that surpass multilingual systems for code-switched texts. These results are confirmed in an alternative task aimed at providing contextual translations for a L2 writing assistant.
Published: 2021

27. The Role of the Auditory and Visual Modalities in the Perceptual Identification of Brazilian Portuguese Statements and Echo Questions

Author: Marc Swerts, Albert Rilliard, João Antônio de Moraes, Luma da Silva Miranda, Tilburg University [Netherlands], Universidade Federal do Rio de Janeiro (UFRJ), Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), and Language, Communication and Cognition
Subjects: Adult, Male, 050101 languages & linguistics, Linguistics and Language, Sociology and Political Science, Statement (logic), media_common.quotation_subject, Audiovisual perception, Prosody, Language and Linguistics, 030507 speech-language pathology & audiology, 03 medical and health sciences, Speech and Hearing, Brazilian Portuguese, Phonetics, Perception, Humans, CUES, 0501 psychology and cognitive sciences, [SHS.LANGUE]Humanities and Social Sciences/Linguistics, media_common, Language, Modalities, 05 social sciences, Echo (computing), General Medicine, SPEECH, PROSODIC PROMINENCE, Echo question, language.human_language, Facial Expression, Identification (information), Acoustic Stimulation, language, Speech Perception, Female, Statement, 0305 other medical science, Psychology, Brazil, Photic Stimulation, Cognitive psychology
Abstract: International audience; This paper presents the results of three perceptual experiments investigating the role of auditory and visual channels for the identification of statements and echo questions in Brazilian Portuguese. Ten Brazilian speakers (five male) were video-recorded (frontal view of the face) while they produced a sentence ("Como você sabe"), either as a statement (meaning "As you know.") or as an echo question (meaning "As you know?"). Experiments were set up with the two different intonation contours. Stimuli were presented in conditions with clear and degraded audio as well as congruent and incongruent information from both channels. Results show that Brazilian listeners were able to distinguish statements and questions prosodically and visually, with auditory performances being dominant over visual ones. In noisy conditions, the visual channel improved the interpretation of prosodic cues robustly, while it degraded them in conditions where the visual information is incongruent with the auditory information. This study confirms the previous findings on auditory and visual integration within speech perception, also when applied to prosodic patterns.
Published: 2021

28. Fine phonetic details for discourse marker disambiguation: a corpus-based investigation

Author: Wu, Yaru, Hutin, Mathilde, Vasilescu, Ioana, Lamel, Lori, Adda-Decker, Martine, Degand, Liesbeth, UCL - SSH/ILC/PLIN - Pôle de recherche en linguistique, Centre de recherche inter-langues sur la signification en contexte (CRISCO), Université de Caen Normandie (UNICAEN), Normandie Université (NU)-Normandie Université (NU), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Traitement du Langage Parlé (TLP ), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), LPP - Laboratoire de Phonétique et Phonologie - UMR 7018 (LPP), Université Sorbonne Nouvelle - Paris 3-Centre National de la Recherche Scientifique (CNRS), and Université Catholique de Louvain = Catholic University of Louvain (UCL)
Subjects: french, [SCCO.LING]Cognitive science/Linguistics, phonetic variation, discourse markers, disambiguation
Abstract: Yaru WU : membre associée au LPP/CNRS (UMR 7018, Université Sorbonne-Nouvelle) et au LISN/CNRS (UMR 9015, Université Paris-Saclay), équipe : Traitement du Langage Parlé (TLP).; International audience; In this study we examine phonetic variation of discourse markers in French, using for this purpose the4-hour richly annotated LOCAS-F corpus. Both linguistic factors and stylistic variables are considered:speech style, part-of-speech category, mean phone duration and vowel formant distributions with respect tothe word status. The results show that the use of discourse markers increases with the degree of spontaneity of the speech. Coordinating conjunctions are thepart-of-speech which is most frequently used for discourse markers. Moreover, the mean phone durationtends to be shorter and the vowel space more centralized when words are employed as discourse markers,suggesting that discourse markers undergo hypoarticulation and, more generally, reduction.
Published: 2021

29. Noisy UGC Translation at the Character Level: Revisiting Open-Vocabulary Capabilities and Robustness of Char-Based Models

Author: Núñez, José Carlos Rosales, Wisniewski, Guillaume, Seddah, Djamé, Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Laboratoire de Linguistique Formelle (LLF - UMR7110), Centre National de la Recherche Scientifique (CNRS)-Université Paris Cité (UPCité), Automatic Language Modelling and ANAlysis & Computational Humanities (ALMAnaCH), Inria de Paris, Institut National de Recherche en Informatique et en Automatique (Inria)-Institut National de Recherche en Informatique et en Automatique (Inria), Association for computational linguistics, and ANR-16-CE33-0021,PARSITI,Analyser l'impossible, Traduire l'improbable(2016)
Subjects: [INFO.INFO-TT]Computer Science [cs]/Document and Text Processing, Computer Science - Computation and Language
Abstract: International audience; This work explores the capacities of character-based Neural Machine Translation to translate noisy User-Generated Content (UGC) with a strong focus on exploring the limits of such approaches to handle productive UGC phenomena, which almost by definition, cannot be seen at training time. Within a strict zero-shot scenario, we first study the detrimental impact on translation performance of various user-generated content phenomena on a small annotated dataset we developed, and then show that such models are indeed incapable of handling unknown letters, which leads to catastrophic translation failure once such characters are encountered. We further confirm this behavior with a simple, yet insightful, copy task experiment and highlight the importance of reducing the vocabulary size hyper-parameter to increase the robustness of character-based models for machine translation.
Published: 2021

30. Intégrer un contexte large en traduction automatique neuronale

Author: Abdul Rauf, Sadaf, Yvon, François, Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), LIMSI-CNRS, and Yvon, François
Subjects: [INFO.INFO-AI] Computer Science [cs]/Artificial Intelligence [cs.AI], [INFO.INFO-TT]Computer Science [cs]/Document and Text Processing, [INFO.INFO-TT] Computer Science [cs]/Document and Text Processing, [INFO.INFO-AI]Computer Science [cs]/Artificial Intelligence [cs.AI]
Published: 2020

31. Priming Neural Machine Translation

Author: Pham, Minh Quang, Xu, Jitao, Crego, Josep-Maria, Senellart, Jean, Yvon, François, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), SYSTRAN, and Publications, Limsi
Subjects: [INFO.INFO-CL] Computer Science [cs]/Computation and Language [cs.CL], translation memories, [INFO]Computer Science [cs], [INFO] Computer Science [cs], Machine translation, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL]
Abstract: International audience; Priming is a well known and studied psychology phenomenon based on the prior presentation of one stimulus (cue) to influence the processing of a response. In this paper, we propose a framework to mimic the process of priming in the context of neural machine translation (NMT). We evaluate the effect of using similar translations as priming cues on the NMT network. We propose a method to inject priming cues into the NMT networkand compare our framework to other mechanisms that perform micro-adaptation during inference. Overall, experiments conducted in a multi-domain setting confirm that adding priming cues in the NMT decoder can go a long way towards improving the translation accuracy.Besides, we show the suitability of our framework to gather valuable information for an NMT network from monolingual resources.
Published: 2020

32. Generative latent neural models for automatic word alignment

Author: Anh Khoa NGO HO, François Yvon, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Université Paris Saclay, and NGO HO, Anh Khoa
Subjects: [INFO.INFO-AI] Computer Science [cs]/Artificial Intelligence [cs.AI], [INFO]Computer Science [cs], [INFO] Computer Science [cs], [INFO.INFO-AI]Computer Science [cs]/Artificial Intelligence [cs.AI]
Abstract: International audience; Word alignments identify translational correspondences between words in a parallel sentence pair and are used, for instance, to learn bilingual dictionaries, to train statistical machine translation systems or to perform quality estimation. Variational autoencoders have been recently used in various of natural language processing to learn in an unsupervised way latent representations that are useful for language generation tasks. In this paper, we study these models for the task of word alignment and propose and assess several evolutions of a vanilla variational autoencoders. We demonstrate that these techniques can yield competitive results as compared to Giza++ and to a strong neural network alignment system for two language pairs.
Published: 2020

33. SimAlign: High Quality Word Alignments Without Parallel Training Data Using Static and Contextualized Embeddings

Author: Jalili Sabet, Masoud, Dufter, Philipp, Yvon, François, Schütze, Hinrich, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Association for Computational Linguistics, and Publications, Limsi
Subjects: FOS: Computer and information sciences, Computer Science - Computation and Language, contextualized embeddings, 02 engineering and technology, [INFO] Computer Science [cs], [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], machine translation, 03 medical and health sciences, 0302 clinical medicine, [INFO.INFO-CL] Computer Science [cs]/Computation and Language [cs.CL], 030221 ophthalmology & optometry, 0202 electrical engineering, electronic engineering, information engineering, [INFO]Computer Science [cs], 020201 artificial intelligence & image processing, Computation and Language (cs.CL), word alignement
Abstract: Word alignments are useful for tasks like statistical and neural machine translation (NMT) and cross-lingual annotation projection. Statistical word aligners perform well, as do methods that extract alignments jointly with translations in NMT. However, most approaches require parallel training data, and quality decreases as less training data is available. We propose word alignment methods that require no parallel data. The key idea is to leverage multilingual word embeddings, both static and contextualized, for word alignment. Our multilingual embeddings are created from monolingual data only without relying on any parallel data or dictionaries. We find that alignments created from embeddings are superior for four and comparable for two language pairs compared to those produced by traditional statistical aligners, even with abundant parallel data; e.g., contextualized embeddings achieve a word alignment F1 for English-German that is 5 percentage points higher than eflomal, a high-quality statistical aligner, trained on 100k parallel sentences., Comment: EMNLP (Findings) 2020
Published: 2020

34. Les parlers romans dans l'atlas sonore des langues et dialectes de Belgique

Author: Boula De Mareüil, Philippe, Mahin, Lucien, Vernier, Frédéric, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Cognition, Perception et Usages (CPU), and Publications, Limsi
Subjects: Walloon, endangered languages, [INFO.INFO-DB]Computer Science [cs]/Databases [cs.DB], Picard, wallon, dialectologie, [SHS.LANGUE] Humanities and Social Sciences/Linguistics, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], dialectology, [INFO.INFO-CL] Computer Science [cs]/Computation and Language [cs.CL], speaking atlas, [INFO.INFO-DB] Computer Science [cs]/Databases [cs.DB], langues en danger, [SHS.LANGUE]Humanities and Social Sciences/Linguistics, linguistique
Abstract: We here describe a speaking linguistic atlas which takes the form of a website presenting an interactive map of Belgium, where one can click on around fifty survey points to listen to and read the same story in endogenous regional languages (Romance and Germanic languages). We recordedthe Aesop fable “The North Wind and the Sun” (used by the International Phonetic Association to illustrate a number of languages of the world) in order to highlight the richness of Picard and Walloon, in particular., Nous décrivons ici un atlas linguistique sonore qui prend la forme d’un site web présentant une carte interactive de Belgique, où l’on peut cliquer sur une cinquantaine de points d’enquête pour écouter et lire une même histoire en langues régionales endogènes (romanes et germaniques). Nous avons enregistré la fable d’Ésope « La bise et le soleil » (utilisée par l’Association phonétique internationale pour illustrer nombre de langues du monde) dans le but de mettre en valeur la richesse du picard et du wallon, en particulier.
Published: 2020

35. A Study of Residual Adapters for Multi-Domain Neural Machine Translation

Author: Pham, Minh Quang, Crego, Josep-Maria, Yvon, François, Senellart, Jean, Publications, Limsi, SYSTRAN, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), and Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)
Subjects: domain adaptation, [INFO.INFO-CL] Computer Science [cs]/Computation and Language [cs.CL], [INFO]Computer Science [cs], Machine translation, [INFO] Computer Science [cs], [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL]
Abstract: International audience; Domain adaptation is an old and vexing problem for machine translation systems. The most common and successful approach to supervised adaptation is to fine-tune a baseline system with in-domain parallel data. Standard fine-tuning however modifies all the network parameters, which makes this approach computationally costly and prone to overfitting. A recent, lightweight approach, instead augments a baseline model with supplementary (small) adapter layers, keeping the rest of the model unchanged. This has the additional merit to leave the baseline model intact and adaptable to multiple domains. In this paper, we conduct a thorough analysis of the adapter model in the context of a multidomain machine translation task. We contrast multiple implementations of this idea using two language pairs. Our main conclusions are that residual adapters provide a fast and cheap method for supervised multi-domain adaptation; our two variants prove as effective as the original adapter model and open perspective to also make adapted models more robust to label domain errors.
Published: 2020

36. The European Language Technology Landscape in 2020: Language-Centric and Human-Centric AI for Cross-Cultural Communication in Multilingual Europe

Author: Rehm, Georg, Marheinecke, Katrin, Hegele, Stefanie, Piperidis, Stelios, Bontcheva, Kalina, Hajic, Jan, Choukri, Khalid, Vasiljevs, Andrejs, Backfried, Gerhard, Prinz, Christoph, Gómez Pérez, José Manuel, Meertens, Luc, Lukowicz, Paul, Van Genabith, Josef, Lösch, Andrea, Slusallek, Philipp, Irgens, Morten, Gatellier, Patrick, Köhler, Joachim, Le Bars, Laure, Anastasiou, Dimitra, Auksoriute, Albina, Bel, Núria, Branco, António, Budin, Gerhard, Daelemans, Walter, De Smedt, Koenraad, Garabik, Radovan, Gavriilidou, Maria, Gromann, Dagmar, Koeva, Svetla, Krek, Simon, Krstev, Cvetana, Lindén, Krister, Magnini, Bernardo, Odijk, Jan, Ogrodniczuk, Maciej, Rögnvaldsson, Eiríkur, Rosner, Mike, Pedersen, Bolette Sandfort, Skadina, Inguna, Tadíc, Marko, Tufis, Dan, Váradi, Tamás, Vider, Kadri, Way, Andy, Yvon, Francois, LS OZ Taal en spraaktechnologie, ILS LLI, Deutsches Forschungszentrum für Künstliche Intelligenz GmbH = German Research Center for Artificial Intelligence (DFKI), Institute for Language and Speech Processing (ILSP), Department of Computer Sciences [Scheffield], University of Sheffield [Sheffield], Charles University [Prague] (CU), Evaluations and Language resources Distribution Agency (ELDA), ELDA, Tilde (Tilde), TILDE, Université libre de Bruxelles (ULB), Smart Systems Laboratory, Thales Services S.A., THALES, Universiteit Antwerpen [Antwerpen], Department of Computational Linguistics, Institute of Bulgarian, University of Sofia-University of Sofia, Jozef Stefan Institute [Ljubljana] (IJS), université de belgrade, University of Helsinki, Human Language Technology [Trento], Fondazione Bruno Kessler [Trento, Italy] (FBK), Instytut Podstaw Informatyki (IPI PAN), Polska Akademia Nauk = Polish Academy of Sciences (PAN), Romanian Academy [IASI], Romanian Academy of Sciences, Hungarian Academy of Sciences (MTA), Dublin City University [Dublin] (DCU), Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), ELDA/ELRA, Yvon, François, THALES [France], Universiteit Antwerpen = University of Antwerpen [Antwerpen], Софийски университет = Sofia University-Софийски университет = Sofia University, Helsingin yliopisto = Helsingfors universitet = University of Helsinki, Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), and Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)
Subjects: FOS: Computer and information sciences, Computer Science - Computation and Language, infrastructural issues, Computer Science - Artificial Intelligence, multilingualism, [INFO.INFO-TT] Computer Science [cs]/Document and Text Processing, Computer Science - Digital Libraries, policy issues, [INFO.INFO-TT]Computer Science [cs]/Document and Text Processing, Artificial Intelligence (cs.AI), Digital Libraries (cs.DL), infrastructures, Computation and Language (cs.CL), National and international projects
Abstract: Multilingualism is a cultural cornerstone of Europe and firmly anchored in the European treaties including full language equality. However, language barriers impacting business, cross-lingual and cross-cultural communication are still omnipresent. Language Technologies (LTs) are a powerful means to break down these barriers. While the last decade has seen various initiatives that created a multitude of approaches and technologies tailored to Europe's specific needs, there is still an immense level of fragmentation. At the same time, AI has become an increasingly important concept in the European Information and Communication Technology area. For a few years now, AI, including many opportunities, synergies but also misconceptions, has been overshadowing every other topic. We present an overview of the European LT landscape, describing funding programmes, activities, actions and challenges in the different countries with regard to LT, including the current state of play in industry and the LT market. We present a brief overview of the main LT-related activities on the EU level in the last ten years and develop strategic guidance with regard to four key dimensions., Comment: Proceedings of the 12th Language Resources and Evaluation Conference (LREC 2020). To appear
Published: 2020

37. A Hierarchical Subspace Model for Language-Attuned Acoustic Unit Discovery

Author: Murat Saraclar, Lucas Ondel, Jan Cernocky, Lukas Burget, Bolaji Yusuf, Faculty of Information Technology [Brno] (FIT / BUT), Brno University of Technology [Brno] (BUT), Boǧaziçi üniversitesi = Boğaziçi University [Istanbul], Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), BUSIM Speech Group, Department of Electrical and Electronic Engineering [Istanbul], Boǧaziçi üniversitesi = Boğaziçi University [Istanbul]-Boǧaziçi üniversitesi = Boğaziçi University [Istanbul], Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Boğaziçi University [Istanbul], and Boğaziçi University [Istanbul]-Boğaziçi University [Istanbul]
Subjects: FOS: Computer and information sciences, Computer Science - Machine Learning, Sound (cs.SD), Computer science, Speech recognition, TIMIT, 02 engineering and technology, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], Computer Science - Sound, Machine Learning (cs.LG), Set (abstract data type), [INFO.INFO-LG]Computer Science [cs]/Machine Learning [cs.LG], Audio and Speech Processing (eess.AS), 0202 electrical engineering, electronic engineering, information engineering, FOS: Electrical engineering, electronic engineering, information engineering, Cluster analysis, Hidden Markov model, ComputingMilieux_MISCELLANEOUS, Frame (networking), Computer Science::Computation and Language (Computational Linguistics and Natural Language and Speech Processing), Speech processing, ComputingMethodologies_PATTERNRECOGNITION, Computer Science::Sound, Embedding, 020201 artificial intelligence & image processing, Subspace topology, Electrical Engineering and Systems Science - Audio and Speech Processing
Abstract: In this work, we propose a hierarchical subspace model for acoustic unit discovery. In this approach, we frame the task as one of learning embeddings on a low-dimensional phonetic subspace, and simultaneously specify the subspace itself as an embedding on a hyper-subspace. We train the hyper-subspace on a set of transcribed languages and transfer it to the target language. In the target language, we infer both the language and unit embeddings in an unsupervised manner, and in so doing, we simultaneously learn a subspace of units specific to that language and the units that dwell on it. We conduct our experiments on TIMIT and two low-resource languages: Mboshi and Yoruba. Results show that our model outperforms major acoustic unit discovery techniques, both in terms of clustering quality and segmentation accuracy., Comment: Submitted to ICASSP 2021
Published: 2020
Full Text: View/download PDF

38. Boosting Neural Machine Translation with Similar Translations

Author: Xu, Jitao, Crego, Josep-Maria, Senellart, Jean, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), SYSTRAN, and Bourse régionale IDF
Subjects: Neural Machine Translation, Fuzzy Match, Translation Memory, 0202 electrical engineering, electronic engineering, information engineering, [INFO]Computer Science [cs], 020201 artificial intelligence & image processing, 02 engineering and technology, 010501 environmental sciences, 01 natural sciences, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], 0105 earth and related environmental sciences
Abstract: International audience; This paper explores data augmentation methods for training Neural Machine Translation to make use of similar translations, in a comparable way a human translator employs fuzzy matches. In particular, we show how we can simply feed the neural model with information on both source and target sides of the fuzzy matches, we also extend the similarity to include semantically related translations retrieved using distributed sentence representations. We show that translations based on fuzzy matching provide the model with "copy" information while translations based on embedding similarities tend to extend the translation "context". Results indicate that the effect from both similar sentences are adding up to further boost accuracy, are combining naturally with model fine-tuning and are providing dynamic adaptation for unseen translation pairs. Tests on multiple data sets and domains show consistent accuracy improvements. To foster research around these techniques, we also release an Open-Source toolkit with efficient and flexible fuzzy-match implementation.
Published: 2020

39. Neural Baselines for Word Alignment

Author: Anh Khoa NGO HO, François Yvon, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)-Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), and Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)
Subjects: FOS: Computer and information sciences, Computer Science - Machine Learning, Neural Machine Translation, Computer Science - Computation and Language, Machine Translation, Word Alignment, [INFO]Computer Science [cs], Computation and Language (cs.CL), [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], Machine Learning (cs.LG)
Abstract: Word alignments identify translational correspondences between words in a parallel sentence pair and is used, for instance, to learn bilingual dictionaries, to train statistical machine translation systems , or to perform quality estimation. In most areas of natural language processing, neural network models nowadays constitute the preferred approach, a situation that might also apply to word alignment models. In this work, we study and comprehensively evaluate neural models for unsupervised word alignment for four language pairs, contrasting several variants of neural models. We show that in most settings, neural versions of the IBM-1 and hidden Markov models vastly outperform their discrete counterparts. We also analyze typical alignment errors of the baselines that our models overcome to illustrate the benefits-and the limitations-of these new models for morphologically rich languages., Comment: The 16th International Workshop on Spoken Language Translation, Nov 2019, Hong Kong, Hong Kong SAR China
Published: 2019
Full Text: View/download PDF

40. Generic and Specialized Word Embeddings for Multi-Domain Machine Translation

Author: Pham, MinhQuang, Crego, Josep, Yvon, François, Senellart, Jean, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)-Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11), SYSTRAN, and Systran
Subjects: Machine Translation, 0202 electrical engineering, electronic engineering, information engineering, [INFO]Computer Science [cs], 020201 artificial intelligence & image processing, 02 engineering and technology, 010501 environmental sciences, 01 natural sciences, Domain Adaptation, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], 0105 earth and related environmental sciences
Abstract: Supervised machine translation works well when the train and test data are sampled from the same distribution. When this is not the case,adaptationtechniques help ensure that the knowledge learned from out-of-domain texts generalises to in-domain sentences. We study here a related setting,multi-domain adaptation, where the number of domains is potentially large and adapting separately to each domain would waste training resources. Our proposal transposes to neural machine translation the feature expansion technique of (DauméIII, 2007): it isolates domain-agnostic from domain-specific lexical representations, while sharing the most of the network across domains. Our experiments use two architectures and two language pairs: they show that our approach, while simple and computationally inexpensive, outperforms several strong baselines and delivers a multi-domain system that successfully translates texts from diverse sources.
Published: 2019
Full Text: View/download PDF

41. Interactions of voice quality settings

Author: Albert Rilliard, Zuleica Camargo, Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11), International Society of Experimental Linguistics, Traitement du Langage Parlé (TLP ), Laboratoire Interdisciplinaire des Sciences du Numérique (LISN), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Institut National de Recherche en Informatique et en Automatique (Inria)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Sciences et Technologies des Langues (STL), and CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-CentraleSupélec-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)
Subjects: Process management, Computer science, media_common.quotation_subject, phonetics, 01 natural sciences, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], auditory perception, 030507 speech-language pathology & audiology, 03 medical and health sciences, 0103 physical sciences, [INFO]Computer Science [cs], Quality (business), [SHS.LANGUE]Humanities and Social Sciences/Linguistics, voice quality, 0305 other medical science, 010301 acoustics, ComputingMilieux_MISCELLANEOUS, media_common
Abstract: International audience; This paper aims at investigating the possible interactions of the voice quality settings (VQS) related to vocal tract (supralaryngeal) and phonatory (laryngeal) activities in the Vocal Profile Analysis Scheme (VPAS), based on Laver's model. The corpus was composed of semi-spontaneous speech samples and repetitions of key-sentences (designed for voice quality evaluation) produced by 71 speakers. The database was perceptually annotated by means of the VPAS, adapted to Brazilian Portuguese (BP), indicating the vocal tract and phonatory VQS. Hierarchical clustering analysiswas run. Results reinforced the interactions between the vocal tract and phonatory VQS, having the general muscular tension (vocal tract and laryngeal hyper and hypofunction) adjustments explained the main grouping of those VQS.
Published: 2019

42. Cross-Lingual Contextual Word Embeddings Mapping With Multi-Sense Words In Mind

Author: Zhang, Zheng, Yin, Ruiqing, Zhu, Jun, Zweigenbaum, Pierre, Information, Langue Ecrite et Signée (ILES), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS), Traitement du Langage Parlé (TLP), CentraleSupélec, Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)-Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), and Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)
Subjects: FOS: Computer and information sciences, ELMo, [INFO.INFO-TT]Computer Science [cs]/Document and Text Processing, Computer Science - Computation and Language, Contextual word embeddings, Cross-lingual mapping, Computation and Language (cs.CL), [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL]
Abstract: Recent work in cross-lingual contextual word embedding learning cannot handle multi-sense words well. In this work, we explore the characteristics of contextual word embeddings and show the link between contextual word embeddings and word senses. We propose two improving solutions by considering contextual multi-sense word embeddings as noise (removal) and by generating cluster level average anchor embeddings for contextual multi-sense word embeddings (replacement). Experiments show that our solutions can improve the supervised contextual word embeddings alignment for multi-sense words in a microscopic perspective without hurting the macroscopic performance on the bilingual lexicon induction task. For unsupervised alignment, our methods significantly improve the performance on the bilingual lexicon induction task for more than 10 points., 12 pages
Published: 2019

43. How Bad are PoS Tagger in Cross-Corpora Settings? Evaluating Annotation Divergence in the UD Project

Author: Wisniewski, Guillaume, Yvon, François, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)-Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11), Association for Computational Linguistics, ANR-16-CE33-0021,PARSITI,Analyser l'impossible, Traduire l'improbable(2016), Wisniewski, Guillaume, and Analyser l'impossible, Traduire l'improbable - - PARSITI2016 - ANR-16-CE33-0021 - AAPG2016 - VALID
Subjects: [INFO.INFO-TT]Computer Science [cs]/Document and Text Processing, Universal Dependencies, [INFO.INFO-TT] Computer Science [cs]/Document and Text Processing, Part-of-speech Tagging
Abstract: International audience; The performance of Part-of-Speech tagging varies significantly across the treebanks of the Universal Dependencies project. This work points out that these variations may result from divergences between the annotation of train and test sets. We show how the annotation variation principle, introduced by Dickinson and Meurers (2003) to automatically detect errors in gold standard, can be used to identify inconsistencies between annotations; we also evaluate their impact on prediction performance.
Published: 2019

44. Measuring text readability with machine comprehension: a pilot study

Author: Marc Benzahra, François Yvon, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)-Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11), and Publications, Limsi
Subjects: 050101 languages & linguistics, Computer science, media_common.quotation_subject, [INFO] Computer Science [cs], 010501 environmental sciences, Machine comprehension, computer.software_genre, 01 natural sciences, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], Correlation, Reading (process), [INFO]Computer Science [cs], 0501 psychology and cognitive sciences, 0105 earth and related environmental sciences, media_common, business.industry, Neural language models, 05 social sciences, Readability, Test (assessment), Comprehension, [INFO.INFO-CL] Computer Science [cs]/Computation and Language [cs.CL], Artificial intelligence, business, computer, On Language, Text readability, Natural language processing
Abstract: International audience; This article studies the relationship between text readability indice and automatic machine understanding systems. Our hypothesis is that the simpler a text is, the better it should be understood by a machine. We thus expect to a strong correlation between readability levels on the one hand, and performance of automatic reading systems on the other hand. We test this hypothesis with several understanding systems based on language models of varying strengths, measuring this correlation on two corpora of journalistic texts. Our results suggest that this correlation is rather small that existing comprehension systems are far to reproduce the gradual improvement of their performance on texts of decreasing complexity.
Published: 2019

45. Using Monolingual Data in Neural Machine Translation: a Systematic Study

Author: Franck Burlot, François Yvon, Lingua Custodia, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)-Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), and Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)
Subjects: FOS: Computer and information sciences, language modeling, Phrase, Machine translation, Data simulation, Computer science, 02 engineering and technology, 010501 environmental sciences, computer.software_genre, 01 natural sciences, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], 0202 electrical engineering, electronic engineering, information engineering, [INFO]Computer Science [cs], 0105 earth and related environmental sciences, Previous generation, Computer Science - Computation and Language, business.industry, Multiple data, 020201 artificial intelligence & image processing, Artificial intelligence, Language model, business, computer, Computation and Language (cs.CL), Natural language processing
Abstract: Neural Machine Translation (MT) has radically changed the way systems are developed. A major difference with the previous generation (Phrase-Based MT) is the way monolingual target data, which often abounds, is used in these two paradigms. While Phrase-Based MT can seamlessly integrate very large language models trained on billions of sentences, the best option for Neural MT developers seems to be the generation of artificial parallel data through \textsl{back-translation} - a technique that fails to fully take advantage of existing datasets. In this paper, we conduct a systematic study of back-translation, comparing alternative uses of monolingual data, as well as multiple data generation procedures. Our findings confirm that back-translation is very effective and give new explanations as to why this is the case. We also introduce new data simulation techniques that are almost as effective, yet much cheaper to implement., Comment: Published in the Proceedings of the Third Conference on Machine Translation (Research Papers), 2018
Published: 2019
Full Text: View/download PDF

46. Controlling Utterance Length in NMT-based Word Segmentation with Attention

Author: Godard, Pierre, Besacier, Laurent, Yvon, François, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris-Sud - Paris 11 (UP11)-Université Paris-Saclay-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Centre National de la Recherche Scientifique (CNRS)-Université Paris Saclay (COmUE)-Université Paris-Sud - Paris 11 (UP11)-Université Paris-Saclay-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Centre National de la Recherche Scientifique (CNRS)-Université Paris Saclay (COmUE), Groupe d’Étude en Traduction Automatique/Traitement Automatisé des Langues et de la Parole (GETALP), Laboratoire d'Informatique de Grenoble (LIG), Université Pierre Mendès France - Grenoble 2 (UPMF)-Université Joseph Fourier - Grenoble 1 (UJF)-Institut polytechnique de Grenoble - Grenoble Institute of Technology (Grenoble INP)-Institut National Polytechnique de Grenoble (INPG)-Centre National de la Recherche Scientifique (CNRS)-Université Grenoble Alpes (UGA)-Université Pierre Mendès France - Grenoble 2 (UPMF)-Université Joseph Fourier - Grenoble 1 (UJF)-Institut polytechnique de Grenoble - Grenoble Institute of Technology (Grenoble INP)-Institut National Polytechnique de Grenoble (INPG)-Centre National de la Recherche Scientifique (CNRS)-Université Grenoble Alpes (UGA), ANR-14-CE35-0002,BULB,Breaking the Unwritten Language Barrier(2014), Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)-Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11), Groupe d’Étude en Traduction Automatique/Traitement Automatisé des Langues et de la Parole (GETALP ), Laboratoire d'Informatique de Grenoble (LIG ), Institut polytechnique de Grenoble - Grenoble Institute of Technology (Grenoble INP )-Centre National de la Recherche Scientifique (CNRS)-Université Grenoble Alpes [2016-2019] (UGA [2016-2019])-Institut polytechnique de Grenoble - Grenoble Institute of Technology (Grenoble INP )-Centre National de la Recherche Scientifique (CNRS)-Université Grenoble Alpes [2016-2019] (UGA [2016-2019]), and ANR-19-P3IA-0003,MIAI,MIAI @ Grenoble Alpes(2019)
Subjects: FOS: Computer and information sciences, Machine Translation, Computer Science - Computation and Language, Computational Language Documentation, Word Segmentation, [INFO]Computer Science [cs], Computation and Language (cs.CL), [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL]
Abstract: One of the basic tasks of computational language documentation (CLD) is to identify word boundaries in an unsegmented phonemic stream. While several unsupervised monolingual word segmentation algorithms exist in the literature, they are challenged in real-world CLD settings by the small amount of available data. A possible remedy is to take advantage of glosses or translation in a foreign, well-resourced, language, which often exist for such data. In this paper, we explore and compare ways to exploit neural machine translation models to perform unsupervised boundary detection with bilingual information, notably introducing a new loss function for jointly learning alignment and segmentation. We experiment with an actual under-resourced language, Mboshi, and show that these techniques can effectively control the output segmentation length., Comment: Accepted to IWSLT 2019 (Hong-Kong)
Published: 2019
Full Text: View/download PDF

47. Adaptor Grammars for the Linguist: Word Segmentation Experiments for Very Low-Resource Languages

Author: Pierre Godard, Annie Rialland, Laurent Besacier, Hélène Maynard, Martine Adda-Decker, Gilles Adda, François Yvon, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)-Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11), Groupe d’Étude en Traduction Automatique/Traitement Automatisé des Langues et de la Parole (GETALP ), Laboratoire d'Informatique de Grenoble (LIG ), Institut polytechnique de Grenoble - Grenoble Institute of Technology (Grenoble INP )-Centre National de la Recherche Scientifique (CNRS)-Université Grenoble Alpes [2016-2019] (UGA [2016-2019])-Institut polytechnique de Grenoble - Grenoble Institute of Technology (Grenoble INP )-Centre National de la Recherche Scientifique (CNRS)-Université Grenoble Alpes [2016-2019] (UGA [2016-2019]), LPP - Laboratoire de Phonétique et Phonologie - UMR 7018 (LPP), and Université Sorbonne Nouvelle - Paris 3-Centre National de la Recherche Scientifique (CNRS)
Subjects: adaptor grammars, Grammar, Computer science, media_common.quotation_subject, Text segmentation, Bantu languages, 02 engineering and technology, Language documentation, Linguistics, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], Constructed language, 03 medical and health sciences, 0302 clinical medicine, Rule-based machine translation, Unsupervised word segmentation, 030221 ophthalmology & optometry, 0202 electrical engineering, electronic engineering, information engineering, 020201 artificial intelligence & image processing, [INFO]Computer Science [cs], On Language, Word (computer architecture), media_common
Abstract: Computational Language Documentation attempts to make the most recent research in speech and language technologies available to linguists working on language preservation and documentation. In this paper, we pursue two main goals along these lines. The first is to improve upon a strong baseline for the unsupervised word discovery task on two very low-resource Bantu languages, taking advantage of the expertise of linguists on these particular languages. The second consists in exploring the Adaptor Grammar framework as a decision and prediction tool for linguists studying a new language. We experiment 162 grammar configurations for each language and show that using Adaptor Grammars for word segmentation enables us to test hypotheses about a language. Specializing a generic grammar with language specific knowledge leads to great improvements for the word discovery task, ultimately achieving a leap of about 30% token F-score from the results of a strong baseline.
Published: 2018

48. Unsupervised Word Segmentation from Speech with Attention

Author: Aline Villavicencio, Laurent Besacier, Lucas Ondel, Alexandre Berard, François Yvon, Marcely Zanon Boito, Pierre Godard, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris-Sud - Paris 11 (UP11)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Université Paris Saclay (COmUE)-Université Paris-Sud - Paris 11 (UP11)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Centre National de la Recherche Scientifique (CNRS)-Université Paris Saclay (COmUE), Laboratoire d'Informatique de Grenoble (LIG ), Institut polytechnique de Grenoble - Grenoble Institute of Technology (Grenoble INP )-Centre National de la Recherche Scientifique (CNRS)-Université Grenoble Alpes [2016-2019] (UGA [2016-2019]), Brno University of Technology [Brno] (BUT), Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 (CRIStAL), Centrale Lille-Université de Lille-Centre National de la Recherche Scientifique (CNRS), School of Computer Science and Electronic Engineering [Essex] (CSEE), University of Essex, Institut Universitaire de France (IUF), Ministère de l'Education nationale, de l’Enseignement supérieur et de la Recherche (M.E.N.E.S.R.), ANR-14-CE35-0002,BULB,Breaking the Unwritten Language Barrier(2014), ANR-11-LABX-0025,PERSYVAL-lab,Systemes et Algorithmes Pervasifs au confluent des mondes physique et numérique(2011), Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)-Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), and Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)
Subjects: FOS: Computer and information sciences, Sequence, Computer Science - Computation and Language, Machine translation, Computer science, Computer Science - Artificial Intelligence, Speech recognition, SIGNAL (programming language), Text segmentation, Bantu languages, 02 engineering and technology, Language documentation, computer.software_genre, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], 030507 speech-language pathology & audiology, 03 medical and health sciences, Artificial Intelligence (cs.AI), 0202 electrical engineering, electronic engineering, information engineering, 020201 artificial intelligence & image processing, 0305 other medical science, computer, Computation and Language (cs.CL), Word (computer architecture)
Abstract: We present a first attempt to perform attentional word segmentation directly from the speech signal, with the final goal to automatically identify lexical units in a low-resource, unwritten language (UL). Our methodology assumes a pairing between recordings in the UL with translations in a well-resourced language. It uses Acoustic Unit Discovery (AUD) to convert speech into a sequence of pseudo-phones that is segmented using neural soft-alignments produced by a neural machine translation model. Evaluation uses an actual Bantu UL, Mboshi; comparisons to monolingual and bilingual baselines illustrate the potential of attentional word segmentation for language documentation., Comment: Interspeech 2018
Published: 2018

49. Automatically Selecting the Best Dependency Annotation Design with Dynamic Oracles

Author: Guillaume Wisniewski, François Yvon, Ophélie Lacroix, Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur ( LIMSI ), Université Paris-Sud - Paris 11 ( UP11 ) -Centre National de la Recherche Scientifique ( CNRS ), Association for Computational Linguistics, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)-Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11), University of Copenhagen = Københavns Universitet (KU), and ANR-16-CE33-0021,PARSITI,Analyser l'impossible, Traduire l'improbable(2016)
Subjects: Dependency (UML), [ INFO ] Computer Science [cs], Computer science, 02 engineering and technology, computer.software_genre, Oracle, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], Set (abstract data type), 03 medical and health sciences, Annotation, 0302 clinical medicine, Dependency grammar, 0202 electrical engineering, electronic engineering, information engineering, [INFO]Computer Science [cs], [ INFO.INFO-CL ] Computer Science [cs]/Computation and Language [cs.CL], Parsing, business.industry, Syntax, dependency parsing, TheoryofComputation_MATHEMATICALLOGICANDFORMALLANGUAGES, 030221 ophthalmology & optometry, 020201 artificial intelligence & image processing, Artificial intelligence, business, computer, Natural language processing, Sentence
Abstract: This work introduces a new strategy to compare the numerous conventions that have been proposed over the years for expressing dependency structures and discover the one for which a parser will achieve the highest parsing performance. Instead of associating each sentence in the training set with a single gold reference, we propose to consider a set of references encoding alternative syntactic representations. Training a parser with a dynamic oracle will then automatically select among all alternatives the reference that will be predicted with the highest accuracy. Experiments on the UD corpora show the validity of this approach.
Published: 2018

50. Fixing Translation Divergences in Parallel Corpora for Neural MT

Author: Jean Senellart, Josep Maria Crego, François Yvon, Minh Quang Pham, Traitement du Langage Parlé (TLP), Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur (LIMSI), Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11)-Université Paris Saclay (COmUE)-Centre National de la Recherche Scientifique (CNRS)-Sorbonne Université - UFR d'Ingénierie (UFR 919), Sorbonne Université (SU)-Sorbonne Université (SU)-Université Paris-Saclay-Université Paris-Sud - Paris 11 (UP11), and SYSTRAN
Subjects: Similarity (geometry), Artificial neural network, Machine translation, Computer science, business.industry, InformationSystems_INFORMATIONSTORAGEANDRETRIEVAL, parallel corpus, 02 engineering and technology, Filter (signal processing), 010501 environmental sciences, computer.software_genre, Translation (geometry), 01 natural sciences, [INFO.INFO-CL]Computer Science [cs]/Computation and Language [cs.CL], Parallel corpora, ComputingMethodologies_DOCUMENTANDTEXTPROCESSING, 0202 electrical engineering, electronic engineering, information engineering, [INFO]Computer Science [cs], 020201 artificial intelligence & image processing, Artificial intelligence, business, computer, Natural language processing, 0105 earth and related environmental sciences
Abstract: Corpus-based approaches to machine translation rely on the availability of clean parallel corpora. Such resources are scarce, and because of the automatic processes involved in their preparation, they are often noisy. This paper describes an unsupervised method for detecting translation divergences in parallel sentences. We rely on a neural network that computes cross-lingual sentence similarity scores, which are then used to effectively filter out divergent translations. Furthermore, similarity scores predicted by the network are used to identify and fix some partial divergences, yielding additional parallel segments. We evaluate these methods for English-French and English-German machine translation tasks, and show that using filtered/corrected corpora actually improves MT performance.
Published: 2018

Catalog

Books, media, physical & digital resources

See catalog results

Searchworks

Select search scope, currently: Articles Catalog books, media & more in Jio Institute collections Articles journal articles & other e-resources

Search

Search Constraints

Refine your results

Search Limiters

Topic

Publication Year Range

Language

Journal

Database

Publisher

57 results on '"Traitement du Langage Parlé (TLP)"'

Search Results

Catalog

Select search scope, currently: Articles

Catalog

books, media & more in Jio Institute collections

Articles

journal articles & other e-resources