Connectionist Language Models For Speech Understanding: The Problem Of Word Order Variation (1999)
| Citations: | 2 - 0 self |
BibTeX
@MISC{Schadle99connectionistlanguage,
author = {Igor Schadle and Jean-Yves Antoine and Daniel Memmi},
title = {Connectionist Language Models For Speech Understanding: The Problem Of Word Order Variation},
year = {1999}
}
Bookmark
 |
 |
 |
 |
 |
 |
OpenURL
Abstract
This paper emphasizes the importance of the question of word-order variation in connectionist language modeling. More precisely, it investigates whether recurrent networks can integrate various linguistic constraints to process variable word-order languages. This paper reports three experiments on the understanding of spoken French that suggest that recurrent architectures could apply to the understanding of variable spoken languages. Keywords: neural networks, language modeling, wordorder, speech understanding. 1. INTRODUCTION: THE PROBLEM OF WORD ORDER VARIATION The use of connectionist methods for speech technologies has so far concerned essentially speech recognition, where the hybridization of Hidden Markov Models (HMM) with neural networks has led to significant results [1]. In spite of some encouraging studies [2,3], connectionist language modeling for spoken dialog systems remains unfortunately widely unsolved. It was claimed for a long time that connectionist architecture...
