Research on a Topic-Based Chinese Language Model

A topic based language model effectively solves the problem of cross domain application of a statistical language model There exist two questions, how to cluster the corpus to different topics and how to combine the topic specific language models First, a new method is adopted to cluster the corpus that has overcome some limitations of the old one Second, an improved algorithm is proposed to combine different language models Not only has the new method improved the performance, but also accelerated the model