Topic Feature Extention Based On Wikipedia

• 1、The school of computer and information technology, Beijing Jiaotong University, Beijing 100044

Abstract：Text classification is an important research in NLP and plays an important role in text mining and information retrieval. Since the curse of dimensionality in traditional Vector Space Model and the limitation of semantic feature representation based on LDA topic model, we propose a topic feature extension method based on Wikipedia to achieve dimensionality reduction and strengthen the text feature expression. By extending the text feature at topic feature level, we can not only reduce the dimensionality of text features to get the classification efficiency improved, but also improve the effect of text classification. Finally, we conduct some classification experiments on the 20Newsgroup and NSF datasets to show that this method has better classification effect.

Keywords： Computer Application Technology Text classification LDA Wikipedia Feature extention