Ajout de nouveaux noms propres au vocabulaire d'un système de transcription en utilisant un corpus diachronique

Proper names are usually keys to understand the information contained in a document. Our work focuses on increasing the vocabulary size of a speech transcription system by automatically retrieving proper names from contemporary diachronic text corpus. We assume that some proper names appear in documents relating to the same time period and in similar lexical contexts. We proposed methods that dynamically augment the automatic speech recognition system vocabulary using lexical and temporal features. Three proposed selection methods are based on co-occurrences statistics inside windows of fixed size, on mutual information and on vector space model. Different metrics for proper name selection in order to limit the vocabulary augmentation are studied. Recognition results show a significant reduction of the proper name error rate using augmented vocabulary with retrieved proper names. MOTS-CLÉS : reconnaissance de la parole, mots hors vocabulaire, noms propres, augmentation du vocabulaire.