Neural Networks for Spoken Language Understanding - Archive ouverte HAL Access content directly
Conference Papers Year : 2017

Neural Networks for Spoken Language Understanding

Abstract

This article explores the use of neural networks, a supervised classification method, to perform a task of spoken language understanding in order to treat automatically user requests over the phone. First the spoken language understanding is defined inside the process chain of automatically treating a user request going from recording the request to giving the appropriate answer. Then are described the several architectures of neural networks used in this study, going from the simplest to the most complex. Some structures allow the network to make past and future prediction in the sentence and other to focus on the important part in a sentence. The results improve significantly as we build more elaborated structure of neural networks. The learning of the neural networks can also be enhanced by bringing additional semantic and syntactic information to the words in the input. Finally we see how the errors made in the other modules involved in the speech processing make the spoken language understanding task more complicated.
Fichier principal
Vignette du fichier
JDOC2017_paper_23.pdf (283.4 Ko) Télécharger le fichier
Origin : Files produced by the author(s)
Loading...

Dates and versions

hal-02997012 , version 1 (09-11-2020)

Identifiers

  • HAL Id : hal-02997012 , version 1

Cite

Edwin Simonnet. Neural Networks for Spoken Language Understanding. 17ème Journée des Doctorants de l’ED STIM, JDOC 2017, May 2017, Nantes, France. ⟨hal-02997012⟩

Collections

UNIV-LEMANS LIUM
31 View
61 Download

Share

Gmail Facebook X LinkedIn More