Using Zero Anaphora Resolution to Improve Text Categorization

Using Zero Anaphora Resolution to Improve Text Categorization. Ching-Long Yeh and Yi-Chun Chen Department of Computer Science and Engineering Tatung University 40 Chungshan N. Rd. 3rd. Section Taipei 104 Taiwan chingyeh@cse.ttu.edu.tw, d8806005@mail.ttu.edu.tw Abstract In Chinese, anaphors are frequently omitted, termed zero anaphor (ZA), from text due to their prominence. Thus the information carried by ZAs in text can not be used to contribute the calculation of text categorization. In this paper, we employ a ZA resolution method to recover the omissions of anaphors in text. Then the resulting text is used as the input of a text categorization system. The experiment result shows that ZA resolution method enhances the accuracy of text categorization from 79% to 84%. 1� Introduction In Chinese text, anaphoric expressions are frequently eliminated, termed zero anaphor (ZA) hereafter, due to their prominence in discourse (Li and Thompson, 1981). Zero anaphors are generally noun phrases that are understood from the context and do not need to be specified. Anaphors in Chinese can be classified as zero, pronominal and nominal forms, as exemplified in (1) by� fl` and #13 respectively'. Zero anaphors are generally noun phrases that are understood from the context and do not need to be specified. ( 1 ) a. WE�r-r,J �i , Zhangsan frightened and ran outside. b. co fiPJ —M (He) bumped into a person. c. VAJ .#r) A in 1--ou He saw clearly that person's appearance. -6=0±', #r) (He) recognized that man as Lisi. The methods of anaphora resolution can be classified into traditional and alternative approaches. The former integrates different knowledge sources or factors (e.g. gender and number agreement, c-command constraints, semantic information) that discount unlikely candidates until a minimal set of plausible candidates is obtained (Grosz et al., 1995; Lappin and Leass, 1994; Okumura and Tamura, 1996; Walker et al., 1998; Yeh and Chen, 2001). Anaphoric relations between anaphors and their antecedents are identified based on the integration of linguistic and domain knowledge. However, it is very labor-intensive and time-consuming to construct a domain knowledge base. The latter employs statistical models or AI techniques, such as machine learning, to compute the most likely candidate This research was supported by the Taiwan National Science Council under Contract No. NSC92-2422-H-036-322 2 We use a coa to denote a zero anaphor, where the superscript a is the index of the antecedent. Also note that superscripts attached to NPs and pronouns are used to represent the indices of their antecedents. 423 (Aone and Bennett, 1995; Connoly et al., 1994; Ge et al., 1998; Seki et al., 2002). This approach can sort out the above problems. However, it heavily relies upon the availability of sufficiently large text corpora that are tagged, in particular, with referential information (Stuckardt, 2002). A recent trend is in search of inexpensive, fast and reliable procedures for anaphora resolution (Baldwin, 1997; Ferrândez et al., 1998; Kennedy and Boguraev, 1996; Mitkov, 1998). The approach relies on cheaper and more reliable NLP tools such as part-of-speech (PUS) tagger and shallow parsers. Without complex …

Content truncated to 500 words