Automatic Sense Derivation for Determinative-Measure Compounds under the Framework of E-HowNet

Computational Linguistics and Chinese Language Processing Vol. 14, No. 1, March 2009, pp. 19-44 19 © The Association for Computational Linguistics and Chinese Language Processing [Received November 3, 2008; Revised May 27, 2009; Accepted May 28, 2009] Automatic Sense Derivation for Determinative-Measure Compounds under the Framework of E-HowNet Chia-Hung Tai∗, Jia-Zen Fan∗, Shu-Ling Huang∗,+, and Keh-Jiann Chen∗ 1B3BAbstract In this paper, we take Determinative-Measure Compounds as an example to demonstrate how the E-HowNet semantic composition mechanism works in deriving the sense representation for a newly coined determinative-measure (DM) compound. First, we define the sense of a closed set of each individual determiner and measure word in E-HowNet representation exhaustively. Afterwards, we make semantic composition rules to produce candidate sense representations for a newly coined DM. Then, we review development set to design sense disambiguation rules. We use these heuristic disambiguation rules to determine the appropriate context-dependent sense of a DM and its E-HowNet representation. The experiment shows that the current system reaches 89% accuracy in DM sense derivation and disambiguation. Keywords: Semantic Composition, Determinative-Measure Compounds, Sense Representations, Extended How Net, How Net 1. Introduction Building a knowledge base is time consuming work. The CKIP Chinese Lexical Knowledge Base has about 80 thousand lexical entries, and their senses are defined in terms of the E-HowNet format. E-HowNet is a lexical knowledge representation system. It extends the framework of HowNet (Dong et al., 2006) to allow semantic composition. Based on the framework of E-HowNet, we intend to establish an automatic semantic composition mechanism to derive sense of compounds and phrases from lexical senses (Chen et al., 2005b), ∗ CKIP, Institute of Information Science, Academia Sinica E-mail: {glaxy; kitajava; kchen} @iis.sinica.edu.tw + Department of Language and Literature Studies, National Hsinchu University of Education E-mail: slhuang@mail.nhcue.edu.tw 20 Chia-Hung Tai et al. (Huang et al., 2008). Determinative-Measure compounds (abbreviated as DM) are the most common compounds in Chinese. As a determiner and a measure normally coin a compound with unlimited versatility, the CKIP group does not define the E-HowNet representations for all DM compounds. Nevertheless, construction patterns for DMs are regular (Li et al., 2006). Therefore, an automatic identification schema in regular expression (Li et al., 2006) and a semantic composition method under the framework of E-HowNet for DM compounds were developed. In this paper, we take DMs as an example to demonstrate how the E-HowNet semantic composition mechanism works in deriving the sense representations for all DM compounds. The remainder of this paper is organized as follows. Section 2 presents the background knowledge of DM compounds and sense representation in E-HowNet. We’ll describe our method in Section 3 and discuss the experiment result in Section 4 before we present conclusions in Section 5. 2. Background There are numerous studies on determiners as well as measures, especially on the types of measures1 F Tai (1994) asserts that classifiers and measures words are often treated together under one single framework of analysis. Chao (1968) treats classifiers as one kind of measure word. In his definition, a …

Content truncated to 500 words