An excerpt from the video of 4 May 2010 session. The debate on Grenelle II de l'environnement has been choosen on the ground of the major controversy it triggered. Green parliament member Yves Cochet did an intervention in which we selected 4 minutes 50 seconds. ; Extrait de la vidéo de la 2e séance du 4 mai 2010. Le débat sur le « Grenelle II de l'environnement » a été sélectionné en raison de la controverse importante qu'il a déclenchée. Le député Vert Yves Cochet y fait une intervention, de laquelle nous avons retenu 4 minutes 50 secondes. ; [1] grenelleII-subpart2.flv (60 Mb) unarchived public 2011-03-21 12:03:54 [2] grenelleII-subpart2.mp4 (32 Mb) public 2011-03-24 14:29:56 [3] grenelleII-subpart2.wav (23 Mb) public 2011-03-21 12:04:20 ; Discourse type: oratory
International audience ; Contrariwise to controlled speech, for which speaker's intention are very limited, spontaneous speech refers to a freer but also richer activity that is characteristic of language interaction.,. Many phenomena appear like hesitations, truncated words, phonetic reductions, etc. In this paper, we first propose a frequency survey of 3 paralinguistic events ("uh", laughter, noise), in different spontaneous corpora: political debate, interviews, task-oriented dialog, informal dialog with instructions and informal dialog without instructions. Even if these events are frequently produced by the speakers, we observe significant differences according to the corpora. For illustrative purposes, we finally show that the results of forced-alignment are significantly improved when the acoustic model of the system includes these events. ; Contrairement à la parole contrôlée, dans laquelle les intentions du locuteur sont très restreintes, la parole spontanée fait référence à une activité plus libre mais aussi plus riche de facteurs caractéristiques de l'interaction langagière. A ce titre, de nombreux phénomènes apparaissent comme les hésitations, les mots tronqués, les réductions phonétiques, etc. Nous proposons dans un premier temps un recensement de 3 événements paralinguistiques ("euh", rire, bruit), dans différents corpus spontanés : débat politique, narration, dialogue orienté tâche, dialogue informel avec consigne et dialogue informel sans consigne. Bien que ces événements soient fréquemment produits par les locuteurs, nous observons des différences significatives selon les corpus. A titre applicatif, nous montrons que les résultats de l'alignement forcé peuvent être nettement améliorés lorsque le système dispose d'un modèle acoustique qui inclut ces événements.
International audience ; Contrariwise to controlled speech, for which speaker's intention are very limited, spontaneous speech refers to a freer but also richer activity that is characteristic of language interaction.,. Many phenomena appear like hesitations, truncated words, phonetic reductions, etc. In this paper, we first propose a frequency survey of 3 paralinguistic events ("uh", laughter, noise), in different spontaneous corpora: political debate, interviews, task-oriented dialog, informal dialog with instructions and informal dialog without instructions. Even if these events are frequently produced by the speakers, we observe significant differences according to the corpora. For illustrative purposes, we finally show that the results of forced-alignment are significantly improved when the acoustic model of the system includes these events. ; Contrairement à la parole contrôlée, dans laquelle les intentions du locuteur sont très restreintes, la parole spontanée fait référence à une activité plus libre mais aussi plus riche de facteurs caractéristiques de l'interaction langagière. A ce titre, de nombreux phénomènes apparaissent comme les hésitations, les mots tronqués, les réductions phonétiques, etc. Nous proposons dans un premier temps un recensement de 3 événements paralinguistiques ("euh", rire, bruit), dans différents corpus spontanés : débat politique, narration, dialogue orienté tâche, dialogue informel avec consigne et dialogue informel sans consigne. Bien que ces événements soient fréquemment produits par les locuteurs, nous observons des différences significatives selon les corpus. A titre applicatif, nous montrons que les résultats de l'alignement forcé peuvent être nettement améliorés lorsque le système dispose d'un modèle acoustique qui inclut ces événements.
International audience ; This paper addresses the problem of the enrichment of transcriptions in the perspective of an automatic phonetization. Phonetization is the process of representing sounds with phonetic signs. There are two general ways to construct a phonetization process: rule based systems (with rules based on inference approaches or proposed by expert linguists) and dictionary based solutions which consist in storing a maximum of phonological knowledge in a lexicon. In both cases, phonetization is based on a manual transcription. Such a transcription is established on the basis of conventions that can differ depending on their working out context. This present study focuses on three different enrichments of such a transcription. Evaluations compare phonetizations obtained from automatic systems to a reference phonetized manually. The test corpus is made of three types of speech in French: conversational speech, read speech and political debate. A specific algorithm for the rule-based system is proposed to deal with enrichments. The final system obtained a phonetization of about 95.2% correct (from 3.7% to 5.6% error rates depending on the corpus).
International audience ; Notre objectif vise à estimer l'influence de différents niveaux d'enrichissement de la transcription sur l'étape de phonétisation de l'oral. Cette étude a été réalisée sur un corpus test de 7 minutes, réparties entre trois types de données différentes (parole conversationnelle spontanée, lecture et discours politique). Les résultats montrent que plus la transcription bénéficie d'enrichissements, meilleure est la phonétisation obtenue, quel que soit le type de corpus.
International audience ; This paper addresses the problem of the enrichment of transcriptions in the perspective of an automatic phonetization. Phonetization is the process of representing sounds with phonetic signs. There are two general ways to construct a phonetization process: rule based systems (with rules based on inference approaches or proposed by expert linguists) and dictionary based solutions which consist in storing a maximum of phonological knowledge in a lexicon. In both cases, phonetization is based on a manual transcription. Such a transcription is established on the basis of conventions that can differ depending on their working out context. This present study focuses on three different enrichments of such a transcription. Evaluations compare phonetizations obtained from automatic systems to a reference phonetized manually. The test corpus is made of three types of speech in French: conversational speech, read speech and political debate. A specific algorithm for the rule-based system is proposed to deal with enrichments. The final system obtained a phonetization of about 95.2% correct (from 3.7% to 5.6% error rates depending on the corpus).
International audience ; This paper addresses the problem of the enrichment of transcriptions in the perspective of an automatic phonetization. Phonetization is the process of representing sounds with phonetic signs. There are two general ways to construct a phonetization process: rule based systems (with rules based on inference approaches or proposed by expert linguists) and dictionary based solutions which consist in storing a maximum of phonological knowledge in a lexicon. In both cases, phonetization is based on a manual transcription. Such a transcription is established on the basis of conventions that can differ depending on their working out context. This present study focuses on three different enrichments of such a transcription. Evaluations compare phonetizations obtained from automatic systems to a reference phonetized manually. The test corpus is made of three types of speech in French: conversational speech, read speech and political debate. A specific algorithm for the rule-based system is proposed to deal with enrichments. The final system obtained a phonetization of about 95.2% correct (from 3.7% to 5.6% error rates depending on the corpus).
International audience ; Notre objectif vise à estimer l'influence de différents niveaux d'enrichissement de la transcription sur l'étape de phonétisation de l''oral. Cette étude a été réalisée sur un corpus test de 7 minutes, réparties entre trois types de données différentes (parole conversationnelle spontanée, lecture et discours politique). Les résultats montrent que plus la transcription bénéficie d'enrichissements, meilleure est la phonétisation obtenue, quel que soit le type de corpus.
International audience ; Notre objectif vise à estimer l'influence de différents niveaux d'enrichissement de la transcription sur l'étape de phonétisation de l'oral. Cette étude a été réalisée sur un corpus test de 7 minutes, réparties entre trois types de données différentes (parole conversationnelle spontanée, lecture et discours politique). Les résultats montrent que plus la transcription bénéficie d'enrichissements, meilleure est la phonétisation obtenue, quel que soit le type de corpus.
An excerpt from the video of 4 May 2010 session. The debate on Grenelle II de l'environnement has been choosen on the ground of the major controversy it triggered. Green parliament member Yves Cochet did an intervention in which we selected 4 minutes at the highest peak of the controversy during which he was interrupted 11 times. ; Extrait de la vidéo de la 2e séance du 4 mai 2010. Le débat sur le « Grenelle II de l'environnement » a été sélectionné en raison de la controverse importante qu'il a déclenchée. Le député Vert Yves Cochet y fait une intervention, de laquelle nous avons retenu 4 minutes du moment le plus vif de la controverse, où le député est interrompu à 11 reprises. ; [1] grenelleII-compte-rendu-04052010-AN.html (0.460 Mb) public 2011-01-23 22:41:11 [2] grenelleII-subpart.flv (54 Mb) unarchived public 2011-01-24 11:21:31 [3] grenelleII-subpart.mp4 (27 Mb) public 2011-01-24 12:26:32 [4] grenelleII-subpart.wav (21 Mb) public 2011-01-21 16:05:45 ; Discourse type: oratory
International audience ; This work was conducted to analyze political debates, with a multimodal point of view. Particularly, we focus on the answers produced by a main speakers after he was disrupted. Our approach relies on the annotations of each modality and on their review. We propose a manual categorization of the observed disruptions. We then apply a categorization method to validate the manual one. The difficulty is to deal with multimodality, missing values and uncertainty in the automatic classification system. ; Cet article traite de l'analyse de débats politiques selon une perspective multimodale. Nous étudions plus particulièrement les réponses aux interruptions lors d'un débat à l'Assemblée nationale. Nous proposons de procéder à l'analyse via des annotations systématiques des différentes modalités. L'analyse argumentative nous a amenée à proposer une typologie de ces réponses. Celle-ci a été mise à l'épreuve d'une classification automatique. La difficulté dans la construction d'un tel système réside dans la nature même des données : multimodales, parfois manquantes et incertaines.
International audience ; The paper introduces an annotation scheme for a political debate dataset which is mainly in the form of video, and audio annotations. The annotation contains various infor- mation ranging from general linguistic to domain specific information. Some are annotated with automatic tools, and some are manually annotated. One of the goals is to use the information to predict the categories of the answers by the speaker to the disruptions. A typology of such answers is proposed and an automatic categorization system based on a multimodal parametrization is successfully performed.
International audience ; This work was conducted to analyze political debates, with a multimodal point of view. Particularly, we focus on the answers produced by a main speakers after he was disrupted. Our approach relies on the annotations of each modality and on their review. We propose a manual categorization of the observed disruptions. We then apply a categorization method to validate the manual one. The difficulty is to deal with multimodality, missing values and uncertainty in the automatic classification system. ; Cet article traite de l'analyse de débats politiques selon une perspective multimodale. Nous étudions plus particulièrement les réponses aux interruptions lors d'un débat à l'Assemblée nationale. Nous proposons de procéder à l'analyse via des annotations systématiques des différentes modalités. L'analyse argumentative nous a amenée à proposer une typologie de ces réponses. Celle-ci a été mise à l'épreuve d'une classification automatique. La difficulté dans la construction d'un tel système réside dans la nature même des données : multimodales, parfois manquantes et incertaines.
International audience ; The paper introduces an annotation scheme for a political debate dataset which is mainly in the form of video, and audio annotations. The annotation contains various infor- mation ranging from general linguistic to domain specific information. Some are annotated with automatic tools, and some are manually annotated. One of the goals is to use the information to predict the categories of the answers by the speaker to the disruptions. A typology of such answers is proposed and an automatic categorization system based on a multimodal parametrization is successfully performed.
International audience ; This work was conducted to analyze political debates, with a multimodal point of view. Particularly, we focus on the answers produced by a main speakers after he was disrupted. Our approach relies on the annotations of each modality and on their review. We propose a manual categorization of the observed disruptions. We then apply a categorization method to validate the manual one. The difficulty is to deal with multimodality, missing values and uncertainty in the automatic classification system. ; Cet article traite de l'analyse de débats politiques selon une perspective multimodale. Nous étudions plus particulièrement les réponses aux interruptions lors d'un débat à l'Assemblée nationale. Nous proposons de procéder à l'analyse via des annotations systématiques des différentes modalités. L'analyse argumentative nous a amenée à proposer une typologie de ces réponses. Celle-ci a été mise à l'épreuve d'une classification automatique. La difficulté dans la construction d'un tel système réside dans la nature même des données : multimodales, parfois manquantes et incertaines.
