A Large-scale Lexical Semantic Knowledge-base of Chinese

A Large-scale Lexical Semantic Knowledge-base of Chinese Wang Hui Department of Chinese Studies National University of Singapore Singapore 117570 chswh@nus.edu.sg Yu Shiwen Institute of Computational Linguistics Peking University Beijing 100871, China yusw@pku.edu.cn Abstract The Semantic Knowledge-base of Contemporary Chinese (SKCC) is a large scale Chinese semantic resource developed by the Institute of Computational Linguistics of Peking University. It provides a large amount of semantic information such as semantic hierarchy and collocation features for 66,539 Chinese words and their English counterparts. Its POS and semantic classification represent the latest progress in Chinese linguistics and language engineering. The descriptions of semantic attributes are fairly thorough, comprehensive and authoritative. The main work in this paper is to introduce the outline of SKCC, and establish a multi-level WSD model based on it. The results indicate that the SCK is effective for word sense disambiguation in Chinese and are likely to be important for general NLP. Key words: semantic knowledge-base, word sense disambiguation (WSD), lexical semantics, Chinese language processing. 1 Introduction Semantic resources play an important role in many areas of Natural Language Processing (NLP). The Institute of Computational Linguistics (ICL) of Peking University has been engaged in research and development of the Semantic Knowledge-base of Contemporary Chinese (SKCC) in the last eight years. This lexicon-building project was collaboration with the Institute of Computing Technology, Chinese Academy of Sciences during 1994-1998, and resulted in a machine-readable bilingual lexicon suitable for use with Machine Translation applications, which contained a fairly complete characterization of the semantic classification, valence specifications and collocation properties for 49 thousands Chinese words and their English counterparts [13. Since 2001, the further development of SKCC has been co-conducted by ICL and Chinese Department of Peking University. At present, SKCC has made great progress. Not only is the scale extended to 66,539 entries, but also the quality has been immensely improved. The semantic classification in the updated edition of SKCC is the embodiment of the very latest progress in Chinese linguistics and language engineering, while the semantic descriptions are comprehensive and thorough. It can provide rich lexical semantic information for various NLP applications. * Supported by China National Fundamental Research Program (973) (PN: G1998030507-4 & G1998030507-1). 243 2 Outline of SKCC 2.1 Scale and Structure SKCC consists of one general database and six sub-databases. The general database contains shared attributes of all the 66,539 entries, while the sub-databases provide detailed descriptions of the distinctive semantic attributes associated with the parts of speech (POS). For example, the verb database has 16 attribute fields, noun database and adjective database has 15 attribute fields respectively (see table 1). Database Name Entries Attribute fields Attribute value nouns 38,478 15 576,555 verbs 21,142 16 338,272 adjective 5,577 15 ,� 83,655 pronouns 236 15 3,540 adverbs 997 11 10,967 numerals 109 11 1,199 General 66,539 8 532,312 Total 133,078 91 1,546,500 Table 1 Scale of SKCC All of the six sub-databases can be linked to the general database through four key fields, namely ENTRY, POS, HOMOMORPHISM and SENSE. As a …

Content truncated to 500 words