关键词: 维基百科, 隐含狄利克雷分布, 文本表示, 文本分类

Abstract: Text representation is the basic task in natural language processing. In general, text representation model can build with sufficient text data. While with insufficient data, it can not complete the task in natural language processing. So, it comes up with a new text represent method to overcome the dilemma. It builds the semantic relationship between words using the link in Wikipedia, and enriches the representation with page rank model diffusing the message to other wiki-items. It verifies that this enrichment can raise the precision, recall and F1-measure of the text classification method.

Key words: Wikipedia, Latent Dirichlet Allocation（LDA）, text representation, text classification