On the Use of Structures in Language Models for Dialogue Specific Solutions For Specific Problems

Availability of large corpora for training language models to develop dialogue systems is rare. Fortunately, for specific dialogue application, many sentences follow a limited number of typical patterns. In a language like French, frequent errors are due to homophones.Three paradigms are proposed in this paper to rescore a trellis of hypothesized words. They are based on sentence patterns detected in the most likely sentence hypothesized in a first recognition phase. This motivates the approach proposed in this paper which suggests to rescore a trellis of hypothesized words based on different types of LMs obtained by adapting to this specific problem some learning methods developed in Artificial Intelligence and Pattern Recognition. These methods are inspired by paradigms known as learning by analogy, explanation-based learning, error correcting parsing and semantic classification. The proposed