An Automated Research Paper Classification Method for the IPC system with the Concept Base
Takanori Shimano, Takashi Yukawa · 2008
In the present paper, a classification method using the Concept Base is proposed and evaluated in the Patent Mining Task of the NTCIR-7 workshop. In this task, research papers are classified into the Interna-tional Patent Classification (IPC) system. The clas-sification enables research papers to be located on a patent map. In order to classify a paper, patent docu-ments that are similar to the paper are retrieved, and the paper is classified as the class of the patent docu-ments. The proposed approach can classify a research paper into correct classes when appropriate patent documents are retrieved. However, there is a problem in that patent documents differ from research papers with respect to document characteristics, even if the technical idea of a patent is the same as that of a re-search paper. For example, the terms and document structure used in a patent differ from those of a paper. Therefore, the classification method required for this task requires an approach that addresses this problem. In order to clarify the performance of the naive method of this approach, the authors classified re-search papers using a simple classification method with the Vector Space Model (as the baseline method). Then, in order to solve the problem caused by the term difference, a classification method using the Con-cept Based Vector Space Model (CBVSM) is proposed. The degree of similarity in this method reflects the se-mantics of the words. The Mean Average Precision (MAP) value of the proposed method did not exceed that of the baseline method. The proposed method is classified in the worst group among the participating teams. However, the proposed method provides better average precision (AP) values than the baseline method for 33 % of top-ics. Based on these results, it is suggested that a more effective method could be constructed using a combi-nation of the baseline method and the method using CBVSM. Keywords:patent classification, IPC system,