This paper utilizes the Bayes Information Criteria(BIC) based model selection theory to evaluate the probability of each topic numbers taking.

• 在此基础上实现了文档矩阵的ICA分解,并根据分离的独立分量获得主题的关键词及其权重。