Gestion d'erreurs pour la fiabilisation des retours automatiques en apprentissage de la prosodie d'une langue seconde

The success of future syste ms for computer assisted foreign language learning relies on providing the learner personalized diagnosis and relevant corrections of its pronunciations. After a presentation of the problem of reliable automatic prosodic feedbacks in language learning, we present our work related to the processing of some errors stemming from the learner and from the system itself. The first part deals with the relevant rejection of incorrect entries (for example due to learner’s errors) while being tolerant to non-native speech deviations. The second part focuses on the automatic phonetic segmentation of nonnative speech. A detailed analysis has showed the benefit of taking into account non-native variants, and lead to determining the classes of phonemes whose temporal boundaries are the most accurate and which should be favored in the design of exercises for language learning. MOTS-CLÉS : apprentissage des langues, parole non native, alignement texte-parole, segmentation phonétique, modélisation stochastique de la parole, rejet entrées incorrectes.