Search

Human biases have been shown to influence the performance of models and algorithms in various fields, including Natural Language Processing. While the study of this phenomenon is garnering focus in recent years, the available resources are still relatively scarce, often focusing on different forms or manifestations of biases. The aim of our work is twofold: 1) gather publicly-available datasets and determine how to better combine them to effectively train models in the task of hate speech detection and classification; 2) analyse the main issues with these datasets, such as scarcity, skewed resources, and reliance on non-persistent data. We discuss these issues in tandem with the development of our experiments, in which we show that the combinations of different datasets greatly impact the models' performance., Comment: 31 pages, 15 Tables, 4 Figures

Brazilian Portuguese and European Portuguese are two varieties of the same language and, despite their close similarities, they exhibit several differences. However, there is a significant disproportion in the availability of resources between the two variants, with Brazilian Portuguese having more abundant resources. This inequity can impact the quality of translation services accessible to European Portuguese speakers. To address this issue, we propose the development of a Brazilian Portuguese to European Portuguese translation system, leveraging recent advancements in neural architectures and models. To evaluate the performance of such systems, we manually curated a gold test set comprising 500 sentences across five different topics. Each sentence in the gold test set has two distinct references, facilitating a straightforward evaluation of future translation models. We experimented with various models by fine-tuning existing Large Language Models using parallel data extracted from movie subtitles and TED Talks transcripts in both Brazilian and European Portuguese. Our evaluation involved the use of conventional automatic metrics as well as a human evaluation. In addition, all models were compared against ChatGPT 3.5 Turbo, which currently yields the best results., Comment: 12 pages, 8 tables

Widely used learned metrics for machine translation evaluation, such as COMET and BLEURT, estimate the quality of a translation hypothesis by providing a single sentence-level score. As such, they offer little insight into translation errors (e.g., what are the errors and what is their severity). On the other hand, generative large language models (LLMs) are amplifying the adoption of more granular strategies to evaluation, attempting to detail and categorize translation errors. In this work, we introduce xCOMET, an open-source learned metric designed to bridge the gap between these approaches. xCOMET integrates both sentence-level evaluation and error span detection capabilities, exhibiting state-of-the-art performance across all types of evaluation (sentence-level, system-level, and error span detection). Moreover, it does so while highlighting and categorizing error spans, thus enriching the quality assessment. We also provide a robustness analysis with stress tests, and show that xCOMET is largely capable of identifying localized critical errors and hallucinations., Comment: Work in progress

We present the joint contribution of Unbabel and Instituto Superior T\'ecnico to the WMT 2023 Shared Task on Quality Estimation (QE). Our team participated on all tasks: sentence- and word-level quality prediction (task 1) and fine-grained error span detection (task 2). For all tasks, we build on the COMETKIWI-22 model (Rei et al., 2022b). Our multilingual approaches are ranked first for all tasks, reaching state-of-the-art performance for quality estimation at word-, span- and sentence-level granularity. Compared to the previous state-of-the-art COMETKIWI-22, we show large improvements in correlation with human judgements (up to 10 Spearman points). Moreover, we surpass the second-best multilingual submission to the shared-task with up to 3.8 absolute points.

Fuzzy Fingerprints have been successfully used as an interpretable text classification technique, but, like most other techniques, have been largely surpassed in performance by Large Pre-trained Language Models, such as BERT or RoBERTa. These models deliver state-of-the-art results in several Natural Language Processing tasks, namely Emotion Recognition in Conversations (ERC), but suffer from the lack of interpretability and explainability. In this paper, we propose to combine the two approaches to perform ERC, as a means to obtain simpler and more interpretable Large Language Models-based classifiers. We propose to feed the utterances and their previous conversational turns to a pre-trained RoBERTa, obtaining contextual embedding utterance representations, that are then supplied to an adapted Fuzzy Fingerprint classification module. We validate our approach on the widely used DailyDialog ERC benchmark dataset, in which we obtain state-of-the-art level results using a much lighter model., Comment: FUZZ-IEEE 2023

State of the art models in intent induction require annotated datasets. However, annotating dialogues is time-consuming, laborious and expensive. In this work, we propose a completely unsupervised framework for intent induction within a dialogue. In addition, we show how pre-processing the dialogue corpora can improve results. Finally, we show how to extract the dialogue flows of intentions by investigating the most common sequences. Although we test our work in the MultiWOZ dataset, the fact that this framework requires no prior knowledge make it applicable to any possible use case, making it very relevant to real world customer support applications across industry., Comment: 16 pages, 8 figures

Current signing avatars are often described as unnatural as they cannot accurately reproduce all the subtleties of synchronized body behaviors of a human signer. In this paper, we propose a new dynamic approach for transitions between signs, focusing on mouthing animations for Portuguese Sign Language. Although native signers preferred animations with dynamic transitions, we did not find significant differences in comprehension and perceived naturalness scores. On the other hand, we show that including mouthing behaviors improved comprehension and perceived naturalness for novice sign language learners. Results have implications in computational linguistics, human-computer interaction, and synthetic animation of signing avatars., Comment: 11 pages, 8 figures

Neural metrics for machine translation evaluation, such as COMET, exhibit significant improvements in their correlation with human judgments, as compared to traditional metrics based on lexical overlap, such as BLEU. Yet, neural metrics are, to a great extent, "black boxes" returning a single sentence-level score without transparency about the decision-making process. In this work, we develop and compare several neural explainability methods and demonstrate their effectiveness for interpreting state-of-the-art fine-tuned neural metrics. Our study reveals that these metrics leverage token-level information that can be directly attributed to translation errors, as assessed through comparison of token-level neural saliency maps with Multidimensional Quality Metrics (MQM) annotations and with synthetically-generated critical translation errors. To ease future research, we release our code at: https://github.com/Unbabel/COMET/tree/explainable-metrics., Comment: Accepted at ACL 2023

Question Generation (QG) is a task of Natural Language Processing (NLP) that aims at automatically generating questions from text. Many applications can benefit from automatically generated questions, but often it is necessary to curate those questions, either by selecting or editing them. This task is informative on its own, but it is typically done post-generation, and, thus, the effort is wasted. In addition, most existing systems cannot incorporate this feedback back into them easily. In this work, we present a system, GEN, that learns from such (implicit) feedback. Following a pattern-based approach, it takes as input a small set of sentence/question pairs and creates patterns which are then applied to new unseen sentences. Each generated question, after being corrected by the user, is used as a new seed in the next iteration, so more patterns are created each time. We also take advantage of the corrections made by the user to score the patterns and therefore rank the generated questions. Results show that GEN is able to improve by learning from both levels of implicit feedback when compared to the version with no learning, considering the top 5, 10, and 20 questions. Improvements go up from 10%, depending on the metric and strategy used., Comment: 27 pages, 8 figures

Recent approaches have attempted to personalize dialogue systems by leveraging profile information into models. However, this knowledge is scarce and difficult to obtain, which makes the extraction/generation of profile information from dialogues a fundamental asset. To surpass this limitation, we introduce the Profile Generation Task (PGTask). We contribute with a new dataset for this problem, comprising profile sentences aligned with related utterances, extracted from a corpus of dialogues. Furthermore, using state-of-the-art methods, we provide a benchmark for profile generation on this novel dataset. Our experiments disclose the challenges of profile generation, and we hope that this introduces a new research direction., Comment: Accepted at SIGDIAL 2023, 4 pages, 2 figures

In this paper, we investigate the problem of including relevant information as context in open-domain dialogue systems. Most models struggle to identify and incorporate important knowledge from dialogues and simply use the entire turns as context, which increases the size of the input fed to the model with unnecessary information. Additionally, due to the input size limitation of a few hundred tokens of large pre-trained models, regions of the history are not included and informative parts from the dialogue may be omitted. In order to surpass this problem, we introduce a simple method that substitutes part of the context with a summary instead of the whole history, which increases the ability of models to keep track of all the previous relevant information. We show that the inclusion of a summary may improve the answer generation task and discuss some examples to further understand the system's weaknesses., Comment: 4 pages, 3 figures, accepted at IberSPEECH 2022

We present the joint contribution of IST and Unbabel to the WMT 2022 Shared Task on Quality Estimation (QE). Our team participated on all three subtasks: (i) Sentence and Word-level Quality Prediction; (ii) Explainable QE; and (iii) Critical Error Detection. For all tasks we build on top of the COMET framework, connecting it with the predictor-estimator architecture of OpenKiwi, and equipping it with a word-level sequence tagger and an explanation extractor. Our results suggest that incorporating references during pretraining improves performance across several language pairs on downstream tasks, and that jointly training with sentence and word-level objectives yields a further boost. Furthermore, combining attention and gradient information proved to be the top strategy for extracting good explanations of sentence-level QE models. Overall, our submissions achieved the best results for all three tasks for almost all language pairs by a considerable margin., Comment: WMT 2022 Quality Estimation shared task

In this paper, we propose an end-to-end sentiment-aware conversational agent based on two models: a reply sentiment prediction model, which leverages the context of the dialogue to predict an appropriate sentiment for the agent to express in its reply; and a text generation model, which is conditioned on the predicted sentiment and the context of the dialogue, to produce a reply that is both context and sentiment appropriate. Additionally, we propose to use a sentiment classification model to evaluate the sentiment expressed by the agent during the development of the model. This allows us to evaluate the agent in an automatic way. Both automatic and human evaluation results show that explicitly guiding the text generation model with a pre-defined set of sentences leads to clear improvements, both regarding the expressed sentiment and the quality of the generated text.

Active learning can play an important role in low-resource settings (i.e., where annotated data is scarce), by selecting which instances may be more worthy to annotate. Most active learning approaches for Machine Translation assume the existence of a pool of sentences in a source language, and rely on human annotators to provide translations or post-edits, which can still be costly. In this article, we assume a real world human-in-the-loop scenario in which: (i) the source sentences may not be readily available, but instead arrive in a stream; (ii) the automatic translations receive feedback in the form of a rating, instead of a correct/edited translation, since the human-in-the-loop might be a user looking for a translation, but not be able to provide one. To tackle the challenge of deciding whether each incoming pair source-translations is worthy to query for human feedback, we resort to a number of stream-based active learning query strategies. Moreover, since we not know in advance which query strategy will be the most adequate for a certain language pair and set of Machine Translation models, we propose to dynamically combine multiple strategies using prediction with expert advice. Our experiments show that using active learning allows to converge to the best Machine Translation systems with fewer human interactions. Furthermore, combining multiple strategies using prediction with expert advice often outperforms several individual active learning strategies with even fewer interactions., Comment: Submitted to Computational Linguistics

In conversational question answering, systems must correctly interpret the interconnected interactions and generate knowledgeable answers, which may require the retrieval of relevant information from a background repository. Recent approaches to this problem leverage neural language models, although different alternatives can be considered in terms of modules for (a) representing user questions in context, (b) retrieving the relevant background information, and (c) generating the answer. This work presents a conversational question answering system designed specifically for the Search-Oriented Conversational AI (SCAI) shared task, and reports on a detailed analysis of its question rewriting module. In particular, we considered different variations of the question rewriting module to evaluate the influence on the subsequent components, and performed a careful analysis of the results obtained with the best system configuration. Our system achieved the best performance in the shared task and our analysis emphasizes the importance of the conversation context representation for the overall system performance., Comment: Submitted manuscript (not anonymized) accepted to the 44th European Conference on Information Retrieval (ECIR) 2022. This preprint has not undergone peer review (when applicable) or any post-submission improvements or corrections. The Version of Record of this contribution is published in Advances in Information Retrieval, and is available online at https://doi.org/10.1007/978-3-030-99739-7_23

In Machine Translation, assessing the quality of a large amount of automatic translations can be challenging. Automatic metrics are not reliable when it comes to high performing systems. In addition, resorting to human evaluators can be expensive, especially when evaluating multiple systems. To overcome the latter challenge, we propose a novel application of online learning that, given an ensemble of Machine Translation systems, dynamically converges to the best systems, by taking advantage of the human feedback available. Our experiments on WMT'19 datasets show that our online approach quickly converges to the top-3 ranked systems for the language pairs considered, despite the lack of human feedback for many translations., Comment: Accepted to ACL-IJCNLP 2021 Main Conference (long paper)

Both dialogue systems and chatbots aim at putting into action communication between humans and computers. However, instead of focusing on sophisticated techniques to perform natural language understanding, as the former usually do, chatbots seek to mimic conversation. Since Eliza, the first chatbot ever, developed in 1966, there were many interesting ideas explored by the chatbots' community. Actually, more than just ideas, some chatbots' developers also provide free resources, including tools and large-scale corpora. It is our opinion that this know-how and materials should not be neglected, as they might be put to use in the human-computer communication field (and some authors already do it). Thus, in this paper we present a historical overview of the chatbots' developments, we review what we consider to be the main contributions of this community, and we point to some possible ways of coupling these with current work in the human-computer communication research line., Comment: 22 pages, 1 figure

In this paper we propose a graph-community detection approach to identify cross-document relationships at the topic segment level. Given a set of related documents, we automatically find these relationships by clustering segments with similar content (topics). In this context, we study how different weighting mechanisms influence the discovery of word communities that relate to the different topics found in the documents. Finally, we test different mapping functions to assign topic segments to word communities, determining which topic segments are considered equivalent. By performing this task it is possible to enable efficient multi-document browsing, since when a user finds relevant content in one document we can provide access to similar topics in other documents. We deploy our approach in two different scenarios. One is an educational scenario where equivalence relationships between learning materials need to be found. The other consists of a series of dialogs in a social context where students discuss commonplace topics. Results show that our proposed approach better discovered equivalence relationships in learning material documents and obtained close results in the social speech domain, where the best performing approach was a clustering technique.

We consider the problem of fully unsupervised learning of grammatical (part-of-speech) categories from unlabeled text. The standard maximum-likelihood hidden Markov model for this task performs poorly, because of its weak inductive bias and large model capacity. We address this problem by refining the model and modifying the learning objective to control its capacity via para- metric and non-parametric constraints. Our approach enforces word-category association sparsity, adds morphological and orthographic features, and eliminates hard-to-estimate parameters for rare words. We develop an efficient learning algorithm that is not much more computationally intensive than standard training. We also provide an open-source implementation of the algorithm. Our experiments on five diverse languages (Bulgarian, Danish, English, Portuguese, Spanish) achieve significant improvements compared with previous methods for the same task.

We present an open-domain Question-Answering system that learns to answer questions based on successful past interactions. We follow a pattern-based approach to Answer-Extraction, where (lexico-syntactic) patterns that relate a question to its answer are automatically learned and used to answer future questions. Results show that our approach contributes to the system's best performance when it is conjugated with typical Answer-Extraction strategies. Moreover, it allows the system to learn with the answered questions and to rectify wrong or unsolved past questions., Comment: 5 pages

When developing a conversational agent, there is often an urgent need to have a prototype available in order to test the application with real users. A Wizard of Oz is a possibility, but sometimes the agent should be simply deployed in the environment where it will be used. Here, the agent should be able to capture as many interactions as possible and to understand how people react to failure. In this paper, we focus on the rapid development of a natural language understanding module by non experts. Our approach follows the learning paradigm and sees the process of understanding natural language as a classification problem. We test our module with a conversational agent that answers questions in the art domain. Moreover, we show how our approach can be used by a natural language interface to a cinema database.