Hierarchical topic-sensitive language models for automatic speech recognition
| Author(s) : | Lidia Mangu, |
| Publisher : | N/A |
| Publication Date : | 1997 |
| ISSN : | N/A |
| Abstract : | Language modeling is the attempt to identify regularities in natural language and capture them in a statistical model. Language models are crucial ingredients in automatic speech recognition where a computer is used to convert spoken text into written form. In this paper we introduce a topic-sensitive language model which estimates lexical probabilities conditional on topic through hierarchical tree smoothing. We use perplexity to evaluate the quality of the new models. 1, |