Heuristic-based Korean Coreference Resolution for Information Extraction

Heuristic-based Korean Coreference Resolution for Information Extraction Euisok Chung, Soojong Lim, Bo-Hyun Yun Human Information Processing Dept. Electronics and Telecommunications Research Institute 161, Kajong-Dong, Yusong-Gu, Daejon, 305-350, KOREA { eschung, isj, ybh } @etri.re.kr Tel. 82-42-860-1014, Fax. 82-42-860-4889 Abstract The information extraction is to delimit in advance, as part of the specification of the task, the semantic range of the output and to filter information from large volumes of texts. The most representative word of the document is composed of named entities and pronouns. Therefore, it is important to resolve coreference in order to extract the meaningful information in information extraction. Coreference resolution is to find name entities co-referencing real-world entities in the documents. Results of coreference resolution are used for name entity detection and template generation. This paper presents the heuristic-based approach for coreference resolution in Korean. We constructed the heuristics expanded gradually by using the corpus and derived the salience factors of antecedents as the importance measure in Korean. Our approach consists of antecedents selection and antecedents weighting. We used three kinds of salience factors that are used to weight each antecedent of the anaphor. The experiment result shows 80% precision. 1� Introduction Information extraction(IE) systems take texts containing natural language as input and produce database templates relevant to a particular application. IE system must create templates describing the relevant entities that are reported on. This requires determining when two or more templates describe the same entity, as templates created from conferencing words to be merged. Thus, it is difficult to resolve coference in order to extract more reliable information. Results of coreference resolution are used for the clue of template generation. In coreference resolution, there are two kinds of problems such as anaphora resolution and name aliases recognition. The name aliases could be resolved by lexical pattern matching or synonym dictionary(Huyck, C. (1998). Fukumoto, J., Masui (1998)). However, the anaphora resolution has more complexities of natural language. It has been studied conservatively in the discourse part of natural language processing. Recently, several proposals addressed that using limited knowledge is better than using heavy linguistic and domain knowledge(Lappin, S. and Leass, H. (1994). Baldwin, F. B. (1995). Mitmov, R. (1998)). This paper presents the heuristic-based approach with limited knowledge such as pattern rules, preference rules, and conditional rules. The resolution procedure is to find antecedents and then to evaluate the weight of antecedents with the heuristics. In this paper, we focus on the anaphora resolution. In Korean, an anaphora consists of 'pronoun' and 'demonstrative pronoun + noun phrases' . 2 Related Works Many researches have been performed to solve the problem of coreference resolution. One is the research of anaphora resolution which is based on the discourse theory such as centering theory(Lappin, S. and Leass, H. (1994). Baldwin, F. B. (1995). Mitmov, R. (1998)), another is the information extraction system in order to apply to MUC(message understanding conference)(Huyck, C. (1998). Yangarber, R. and Grishman, R. (1998). Urbanowicz, R. A. and Nettleton, D. J. (1998). Humphreys,K., 50 Gaizauskas, R. …

Content truncated to 500 words