Text Categorization for Authorship based on the Features of Lingual Conceptual Expression

Text Categorization for Authorship based on the Features of Lingual Conceptual Expression* Quan Zhang a, Yun-liang Zhang a,b, and Yi Yuan a a The Institute of Acoustics, CAS, Beijing 100080, P. R. China b Institute of Scientific & Technical Information of China, Beijing 100038, P. R. China {zhq, yuan}@mail.ioa.ac.cn Abstract. The text categorization is an important field for the automatic text information processing. Moreover, the authorship identification of a text can be treated as a special text categorization. This paper adopts the conceptual primitives’ expression based on the Hierarchical Network of Concepts (HNC) theory, which can describe the words meaning in hierarchical symbols, in order to avoid the sparse data shortcoming that is aroused by the natural language surface features in text categorization. The KNN algorithm is used as computing classification element. Then, the experiment has been done on the Chinese text authorship identification. The experiment result gives out that the processing mode that is put forward in this paper achieves high correct rate, so it is feasible for the text authorship identification. Keywords: Lingual Conceptual Space, Hierarchical Network of Concepts (HNC) theory, Text Categorization, KNN Algorithm, Authorship of a text. 1. Introduction Along with the progress of social information industry, especially with the Internet development, more and more documents exist in the electronical form. It provides convenience for information automatic processing. The text categorization is the basic work for the automatic processing, and it is the foundation for the information retrieval, information mining, and question-answer system too. In some applications, it is need to identify the text authorship. The identification need use the documents written by the author as reference. Since the literary style of an author is relatively steady in some time, if we can mine the style character of the author then the identification is realized. Of course, the more documents which are gathered, the better result which is achieved. In many cases, the text authorship identification is treated as text categorization as the researchers can accomplish the work according the text categorization way. In this way, the frequency of glossary, punctuation, n-Gram string, syntax feature, average sentence length, the length of paragraph, and so on, are be used as the features to identify the author of a text * The paper is supported by the National Fundamental Research Project (973 Project) (Grant No. 2004CB318104) and The Knowledge Innovation Engineering Project of The Institute of Acoustics, CAS (Grant No. 0654091431). Copyright 2007 by Quan Zhang, Yun-liang Zhang and Yi Yuan 515 [Burrows J. F. 1987, Baayen R. H. 1996, David I Holmes 1997, Yuta Tsuboi 2002]. The texts are often literature, and sometimes they are internet text, such as E-mail [Olivier de Vel 2001]. Presently, there are a few research works in the Chinese text authorship identification, and many of them orient to the linguistic research for the concrete literature. Sun and Jin [Yi-jian Jin, Xiao-ming Sun, and Shao-ping Ma, 2003] use the vector space model (VSM) which takes the syntax structural words as feature to identify …

Content truncated to 500 words