Search

This paper reports studies on annotating and automatically detecting dialog acts in human-human spoken dialogs. The work reposes on three hypotheses: first, the succession of dialog acts is strongly constrained; second, the initial word and semantic class of word are more important for identifying dialog acts than the complete exact word sequence of an utterance; third, most of the important information is encoded in specific entities. A memory based learning approach is used to detect dialog acts. For each utterance unit, eight dialog acts are systematically annotated. Experiments have been conducted using different levels of information, with and without the use of dialog history information. In order to assess the generality of the method, the specific entity tag based model trained on a French corpus was tested on an English corpus for a similar task and on a French corpus from a different domain. A correct dialog act detection rate of about 86% is obtained for the same domain/language condition and 77% for the cross-language or cross-domain conditions.

Dans ce papier, nous traitons des structures syllabiques et de leur variation dans un corpus de parole en francais issu d'entrevues radio-diffusees. Un des buts est de montrer comment des systemes de reconnaissance automatique de la parole (RAP) peuvent servir d'outils linguistiques pour explorer de facon coherente des corpus virtuellement illimites. Des sous-ensembles automatiquement selectionnes peuvent etre verifies manuellement pour accroitre notre connaissance des variantes de prononciation. Notre conviction est qu'une meilleure formalisation des mecanismes a l'oeuvre dans la parole contribuera en definitive a ameliorer la modelisation des prononciations et les systemes de RAP : cette etude se veut une etape dans cette direction. Les phenomenes linguistiques auxquels nous nous interessons en particulier sont les variantes sequentielles (i.e. celles qui induisent un nombre variable de phonemes), qui peuvent selon les cas conduire a une restructuration syllabique : ces variantes, frequentes en parole spontanee, sont connues pour poser probleme a la reconnaissance. Pour se focaliser sur elles, une methodologie a ete mise au point, utilisant des descriptions aux niveaux phonematique, syllabique et lexical. Cette etude repose sur un corpus de parole de radio constitue de trente emissions d'une heure. La parole spontanee revele une plus grande proportion de syllabes fermees que dans les syllabes canoniques derivees des transcriptions orthographiques : comme on peut s'y attendre, le schwa optionnel contribue pour une grande part a la variation de structure syllabique. Des phenomenes, moins bien decrits ont egalement ete observes : d'autres voyelles telles que /u/, /e/, /i/ et /a/ peuvent tomber en position inaccentuee (non finale). Les syllabes CV non accentuees, precedees d'une syllabe ouverte, sont enclines a la restructuration: effacement de la voyelle et transfert attaque-coda. Les syllabes complexes tendent a etre simplifiees: les consonnes liquides tombent souvent, plus en position de coda qu'en position d'attaque. Le /v/ est la consonne la plus facilement elidee independamment de sa position dans la syllabe. Enfin un pourcentage substantiel de syllabes faibles de fin de mot, ayant un schwa comme noyau, peuvent disparaitre, ainsi que les mots outils monosyllabiques independamment de l'identite de leur noyau vocalique.

In this paper we report on a series of user trials carried out to assess the performance and usability of the Multimodal Multimedia Service Kiosk (MASK) prototype. The aim of the ESPRIT MASK project was to pave the way for advanced public service applications with user interfaces employing multimodal, multimedia input and output. The prototype kiosk was developed after analyzing the technological requirements in the context of users performing travel enquiry tasks, in close collaboration with the French Railways (SNCF) and the Ergonomics group at the University College of London (UCL). The time to complete the transaction with the MASK kiosk is reduced by about 30% compared to that required for the standard kiosk, and the transaction success rate is 85% for novices and 94% once familiar with the system. In addition to meeting or exceeding the performance goals set at the project onset in terms of success rate, transaction time, and user satisfaction, the MASK kiosk was judged to be user-friendly and simple to use.

This paper reports on activites at LIMSI over the last few years directed at the transcription of broadcast news data. We describe our development work in moving from laboratory read speech data to real-world or `found' speech data in preparation for the DARPA evaluations on this task from 1996 to 1999. Two main problems needed to be addressed to deal with the continuous flow of inhomogenous data. These concern the varied acoustic nature of the signal (signal quality, environmental and transmission noise, music) and different linguistic styles (prepared and spontaneous speech on a wide range of topics, spoken by a large variety of speakers). The problem of partitioning the continuous stream of data is addressed using an iterative segmentation and clustering algorithm with Gaussian mixtures. The speech recognizer makes use of continuous density HMMs with Gaussian mixture for acoustic modeling and 4-gram statistics estimated on large text corpora. Word recognition is performed in multiple passes, where current hypotheses are used for cluster-based acoustic model adaptation prior to the next decoding pass. The overall word transcription error of the LIMSI evaluation systems were 27.1% (Nov96, partitioned test data), 18.3% (Nov97, unpartitioned data), 13.6% (Nov98, unpartitioned data) and 17.1% (Fall99, unpartitioned data with computation time under 10× real-time).

Speaker verification has been the subject of active research for many years, yet despite these eAorts and promising results on laboratory data, speaker verification performance over the telephone remains below that required for many applications. This experimental study aimed to quantify speaker recognition performance out of the context of any specific application, as a function of factors more-or-less acknowledged to aAect the accuracy. Some of the issues addressed are: the speaker model (Gaussian mixture models are compared with phone-based models), the influence of the amount and content of training and test data on performance; performance degradation due to model aging and how can this be counteracted by using adaptation techniques; achievable performance levels using text-dependent and text-independent recognition modes. These and other factors were addressed using a large corpus of read and spontaneous speech (over 250 hours collected from 100 target speakers and 1000 imposters) in French, designed and recorded for the purpose of this study. On these data, the lowest equal error rate is 1% for the text-dependent mode when two trials are allowed per verification attempt and with a minimum of 1.5 s of speech per trial. ” 2000 Elsevier Science B.V. All rights reserved.

This contribution aims at evaluating the use of pronunciation variants for different recognition system configurations, languages and speaking styles. This study is limited to the use of variants during speech alignment, given an orthographic transcription of the utterance and a phonemically represented lexicon and is thus focused on the modeling capabilities of the acoustic word models. To measure the need for variants we have defined the variant2+ rate which is the percentage of words in the corpus not aligned with the most common phonemic transcription. This measure may be indicative of the possible need for pronunciation variants in the recognition system. Pronunciation lexica have been automatically created so as to include a large number of variants (overgeneration). In particular, lexica with parallel and sequential variants were automatically generated in order to assess the spectral and temporal modeling accuracy. We first investigated the dependence of the aligned variants on the recognizer configuration. Then a cross-lingual study was carried out for read speech in French and American English using the BREF and the WSJ corpora. A comparison between read and spontaneous speech was made for French based on alignments of BREF (read) and Mask (spontaneous) data. Comparative alignment results using different acoustic model sets demonstrate the dependency between the acoustic model accuracy and the need for pronunciation variants. The alignment results obtained with the above lexica have been used to study the link between word frequencies and variants using different acoustic model sets. Cette contribution vise a evaluer l'utilisation des variantes de prononciation pour differentes configurations de systeme, differentes langues et differents types d'elocution. Cette etude se limite a l'utilisation de variantes pendant l'alignement automatique de la parole etant donnee une transcription orthographique correcte et un lexique de prononciation. Nous focalisons ainsi notre etude sur la capacite des modeles acoustiques des mots a rendre compte du signal observe. Pour evaluer le besoin de variantes nous avons defini le taux de variant2+ qui correspond au pourcentage de mots du corpus qui ne sont pas alignes avec la meilleure transcription phonemique. Ce taux peut etre considere comme indicatif d'un eventuel besoin de variantes de prononciation dans le systeme de reconnaissance. Differents lexiques de prononciation ont ete crees automatiquement generant differents types et quantites de variantes (avec surgeneration). En particulier des lexiques avec des variantes paralleles et sequentielles ont ete distingues afin d'evaluer la precision de la modelisation spectrale et temporelle. Dans une premiere etape nous avons montre le lien entre le besoin de variantes de prononciation et la qualite des modeles acoustiques. Nous avons ensuite compare differents phenomenes de variantes pour l'anglais et le francais sur des grands corpus de parole lue (WSJ et BREF). Une comparaison entre parole spontanee et parole lue est presentee. Cette etude montre que le besoin de variantes diminue avec la precision des modeles acoustiques. Pour le francais, elle permet de reveler l'importance des variantes sequentielles, en particulier du e-muet.

This paper describes the RailTel system developed at LIMSI to provide vocal access to static train timetable information in French, and a field trial carried out to assess the technical adequacy of available speech technology for interactive services. The data collection system used to carry out the field trials is based on the LIMSI Mask spoken language system and runs on a Unix workstation with a high quality telephone interface. The spoken language system allows a mixed-initiative dialog where the user can provide any information at any point in time. Experienced users are thus able to provide all the information needed for database access in a single sentence, whereas less experienced users tend to provide shorter responses, allowing the system to guide them. The RailTel field trial was carried out using a common methodology defined by the consortium. 100 novice subjects participated in the field trials, each calling the system one time and completing a user questionnaire. Of the callers, 72% successfully completed their scenario. The subjective assessment of the prototype was for the most part favourable, with subjects expressing an interest in using such a service.