Seminaire LIA

Vendredi 23/02: Sahar Ghannay



Titre : A study of continuous word representations applied to ASR error detection



Résumé :



Continuous word representations (word embeddings) have proven to be a great asset in various natural language  and spoken language processing tasks.  Many approaches have been proposed to build word embeddings, the most popular ones are based on speech audio, continuous bag of word, syntactic dependency, on co-occurrences matrix, etc.

The major concern of this talk is the study of different types of continuous word representations and their application to ASR error detection task and other tasks.

This study concerns, the evaluation and the combination of linguistic word embeddings, the construction and the evaluation of acoustic embeddings and finally the construction of task specific sentence embeddings.

Finally, we exploited the linguistic and acoustic embeddings as well as the information provided by our ASR error detection system in several downstream applications.



Bio :

Sahar Ghannay is a post-doc researcher at LIUM laboratory (under the direction of Yannick Estève) since October 2017, working on end-to-end neural systems for speech understanding, translation and named entity detection.

She received a PhD in Computer Science, entitled « A study of continuous word representations applied to ASR error detection », from Le Mans University in September 2017.

Sahar did an internship at Apple within the Siri Speech team under the direction of Xiaochuan Niu and Ilya Oparin.

Her main research interests are continuous word representations and their application to natural language and spoken language processing.

Undefined
event_date: 
Vendredi, 23 Février, 2018 - 11:00 to 12:00


Laboratoire Informatique d'Avignon

Université d'Avignon et des Pays de Vaucluse
339 chemin des Meinajaries, Agroparc BP 91228, 84911 Avignon cedex 9
+33 (0)4 90 84 35 00