Language understanding using n-multigram models

Abstract

In this work, we present an approach to language understanding using corpus-based and statistical language models based on multigrams. Assuming that we can assign meanings to segments of words, the n-multigram modelization is a good approach to model sequences of segments that have semantic information associated to them. This approach has been applied to the task of speech understanding in the framework of a dialogue system that answers queries about train timetables in Spanish. Some experimental results are also reported.