Category-based Language Models in a Spanish Spoken Dialogue System

The main goal of this work is to study if a language model based on categories could improve the performance of a dialogue system application as it does when not spontaneous and bigger English corpora are used. Firstly, several sets of categories, which are generated on the basis of different classification criteria, are obtained. Then, for each criterion, two language models are generated: A language model based on category k-grams and a hybrid model that is an interpolation of a word-based language model and a category-based language model. Finally, experiments on a spontaneous dialogue corpus in Spanish are reported. These experiments have been carried out in terms of Perplexity and Word Error Rate.