Journal of South China University of Technology (Natural Science Edition) ›› 2005, Vol. 33 ›› Issue (8): 58-61.

Previous Articles     Next Articles

Patent Categorization Based on Kernel Vector Space Model

Ding Yue-hua  Wen Gui-hua  Guo Wei-qiang   

  1. Research Institute of Computer Applications,South China Univ.of Tech.,Guangzhou 510640,Guangdong,China
  • Received:2004-10-19 Online:2005-08-25 Published:2005-08-25
  • Contact: 丁月华(1950-)，男，高工，主要从事人工智能研究 E-mail:crwqguo@scut.edu.cn
  • About author:丁月华(1950-)，男，高工，主要从事人工智能研究
  • Supported by:

    国家自然科学基金资助项目(60003019)和华南理工大学高水平大学建设项目(159-D65010)

PDF

1006

Abstract:

A novel model,namely,kernel vector space model,is established by using the kernel function to im-prove the vector space.In this model,the Mercer kernel is used to map the data in the original space to a high-di-mensional feature space in which data can be identified as in the vector space. The proposed model is then applied to the patent categorization,with the theoretical and experimental results indicating its greater correctness than the traditional vector space model.

Key words: text categorization, vector space model, kernel function