Make text look like speech: disfluency generation using sequence-to-sequence neural networks

Henri Lasselin 1 Gwénolé Lecorvé 1
1 EXPRESSION - Expressiveness in Human Centered Data/Media
UBS - Université de Bretagne Sud, IRISA-D6 - MEDIA ET INTERACTIONS
Abstract : The synthesis of spontaneous natural speech is a challenge. One way to approach it is to introduce disfluencies since the latter are very present in spontaneous speech. Recently, work has proposed a method to generate disfluencies using language models and conditional random fields. However, neural networks can deal with many problems in natural language processing and it may be wise to use them to produce disfluencies. In this document, we draw up the state of the art of disfluencies as well as of sequence-to-sequence models in order to realize this work during an M.Sc. internship and to follow the most appropriate tracks.
Complete list of metadatas

Cited literature [24 references]  Display  Hide  Download

https://hal.inria.fr/hal-01738344
Contributor : Gwénolé Lecorvé <>
Submitted on : Tuesday, March 20, 2018 - 2:31:06 PM
Last modification on : Thursday, February 7, 2019 - 4:49:27 PM
Long-term archiving on : Tuesday, September 11, 2018 - 9:23:01 AM

File

Make_text_look_like_speech_Vfi...
Files produced by the author(s)

Identifiers

  • HAL Id : hal-01738344, version 1

Citation

Henri Lasselin, Gwénolé Lecorvé. Make text look like speech: disfluency generation using sequence-to-sequence neural networks. [Rapport de recherche] Univ Rennes, CNRS, IRISA, France; IRISA, équipe EXPRESSION. 2018. ⟨hal-01738344⟩

Share

Metrics

Record views

179

Files downloads

117