Xavier Rodet


2008

pdf bib
IrcamCorpusTools : plate-forme pour les corpus de parole [IRCAM Corpus Tools: Managing speech corpora]
Grégory Beller | Christophe Veaux | Gilles Degottex | Nicolas Obin | Pierre Lanchantin | Xavier Rodet
Traitement Automatique des Langues, Volume 49, Numéro 3 : Recherches actuelles en phonologie et en phonétique : interfaces avec le traitement automatique des langues [Current Research in Phonology and Phonetics: Interfaces with Natural-Language Processing]

pdf bib
Automatic Phoneme Segmentation with Relaxed Textual Constraints
Pierre Lanchantin | Andrew C. Morris | Xavier Rodet | Christophe Veaux
Proceedings of the Sixth International Conference on Language Resources and Evaluation (LREC'08)

Speech synthesis by unit selection requires the segmentation of a large single speaker high quality recording. Automatic speech recognition techniques, e.g. Hidden Markov Models (HMM), can be optimised for maximum segmentation accuracy. This paper presents the results of tuning such a phoneme segmentation system. Firstly, using no text transcription, the design of an HMM phoneme recogniser is optimised subject to a phoneme bigram language model. Optimal performance is obtained with triphone models, 7 states per phoneme and 5 Gaussians per state, reaching 94.4% phoneme recognition accuracy with 95.2% of phoneme boundaries within 70 ms of hand labelled boundaries. Secondly, using the textual information modeled by a multi-pronunciation phonetic graph built according to errors found in the first step, the reported phoneme recognition accuracy increases to 96.8% with 96.1% of phoneme boundaries within 70 ms of hand labelled boundaries. Finally, the results from these two segmentation methods based on different phonetic graphs, the evaluation set, the hand labelling and the test procedures are discussed and possible improvements are proposed.

pdf bib
IrcamCorpusTools: an Extensible Platform for Spoken Corpora Exploitation
Christophe Veaux | Gregory Beller | Xavier Rodet
Proceedings of the Sixth International Conference on Language Resources and Evaluation (LREC'08)

Corpus based methods are increasingly used for speech technology applications and for the development of theoretical or computer models of spoken languages. These usages range from unit selection speech synthesis to statistical modeling of speech phenomena like prosody or expressivity. In all cases, these usages require a wide range of tools for corpus creation, labeling, symbolic and acoustic analysis, storage and query. However, if a variety of tools exists for each of these individual tasks, they are rarely integrated into a single platform made available to a large community of researchers. In this paper, we propose IrcamCorpusTools, an open and easily extensible platform for analysis, query and visualization of speech corpora. It is already used for unit selection speech synthesis, for prosody and expressivity studies, and to exploit various corpora of spoken French or other languages.