--------

Modèles Informatiques du Langage et de la Cognition - MILC

Artificial Intelligence Group
Computer Science Department

Français Research Papers Home


François YVON

Ph.D. Student

Biography

Born 1964 in Poitiers, France, François YVON graduated from the Ecole Polytechnique in 1987, and from the Ecole Nationale de Statistique et d'Administration Economique (ENSAE) in 1989. During the same period, he worked as Research Associate in a Social Science Laboratory, where his main research topic was the automatic treatment of open questions in surveys using lexicometric methods. After three years as Project Manager in the Pharmaceutical Industry, he joined TELECOM Paris in 1992 to start a Ph.D in computationnal Linguistics under the direction of Pr. A. BONNET. Since then, his research activity has been devoted to the question of grapheme to phoneme transcription of proper names (people first and last names, place names and acronyms) for speech synthesis. This includes both a careful study of specific morphological, phonological and syntactical properties of these ``weird'' linguistic componants, and a study of self-learning techniques that could help to carry out phonemic transcription.
He is is also responsible for building a 1 million entries lexicon (including a phonemic transcription) of French proper names, within the LRE project ONOMASTICA.

Research

Thesis director : Alain BONNET
Research Group : MILC
Laboratory : Département Informatique, TELECOM Paris
Intended examination date : December 1995

Grapheme-to-Phoneme Conversion Systems for Proper Names

Pronuncing proper names (patronyms, toponyms or acronyms) is commonly known to be a difficult task for many speech synthesis systems. These difficulties have numerous causes, such as exotic etymologies or the vagueness of the concept of "a good pronunciation for a name", etc.
The goal of this PhD work is to explore and/or implement systems (either rule or analogy based) specifically devoted to the grapheme to phoneme conversion of these weird linguistic entities; systems which would be of great interest for the development of name intensive real word text-to-speech applications.

Papers


Last changes November 30 1994
François Yvon (yvon@inf.enst.fr)