Chinese WordNet Domains: Bootstrapping Chinese WordNet with Semantic Domain Labels

Chinese WordNet Domains: Bootstrapping Chinese WordNet with Semantic Domain Labels ∗∗∗∗ Lung-Hao Lee a, Yu-Ting Yu a, and Chu-Ren Huang a,b aInstitute of Linguistics, Academia Sinica, 128 Academia Road, Section 2, Taipei 115, Taiwan {lunghao, tina0822, churen}@gate.sinica.edu.tw bDepartment of Chinese & Bilingual Studies, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong churen.huang@inet.polyu.edu.hk Abstract. We bootstrapped Chinese WordNet with semantic domain labels of WordNet Domains for constructing a language resource called Chinese WordNet Domains. The bootstrapping methods work from three aspects: 1) Princeton WordNet alignment, 2) lexical semantic relations and 3) domain taxonomy mapping. Experimental results of our proposed bootstrapping based domain predication achieve satisfying effects. We believe the resulting Chinese WordNet Domains will be the first oriental language resource, which can be used to interoperate with the existing WordNet Domains of several languages and benefits for cross-language and domain-specific researches and applications. In addition, we also plan to release resulting Chinese WordNet Domains to the community for research purposes. Keywords: Bootstrapping, Chinese WordNet, WordNet Domains, Multi-label. ∗ This work was funded by National Science Council, Taiwan under Grants NSC 97-2923-I-001-001-MY3, and also cooperated with EU-FP7 KYOTO project. Copyright 2009 by Lung-Hao Lee, Yu-Ting Yu, and Chu-Ren Huang 1 Introduction Princeton WordNet is an English lexical database that groups nouns, verbs, adjectives and adverbs into sets of cognitive synonyms, which are named as synsets (Fellbaum, 1998; Miller, 1995). The Global WordNet Association (GWA), built on the results of Princeton WordNet and Euro WordNet (Vossen, 2004), is a free and public association that provides a platform that shares and connects all languages in the world. For Mandarin Chinese in Taiwan, Huang et al. (2004a) constructed the Academia Sinica Bilingual Ontological Wordnet (Sinica BOW), which integrates WordNet, English-Chinese Translation Equivalents Database (ECTED) and SUMO for cross-language linguistic studies. As a follow-up, Chinese WordNet has been built as a robust lexical knowledge system which embodies a precise expression of sense and sense relations as well (Huang et al., 2008b). In recent years, WordNet-like resources have become one of the most reliable and essential resource for linguistic studies for all languages (Niles and Pease, 2003; Budanitsky and Hirst, 2006; Soria et al., 2009a ). Semantic domain labels, characterized by domain-specific lexica, are profitably used to describe texts and word senses according to general subjects, such as sport, finance, and politics. WordNet Domains (Magnini and Cavaglia, 2000) was created by extending the Princeton WordNet with domains labels. Synsets have been semi-automatically annotated with at least one domain label. A domain can include synsets of different part-of-speech and from different WordNet sub-hierarchies. So far the existing WordNets such as Italian WordNet, 288 23rd Pacific Asia Conference on Language, Information and Computation, pages 288–296 Spain WordNet, Hebrew WordNet, and Romanian WordNet are annotated with the same domains labels from WordNet Domains. WordNet Domains has been viewed as an important language resource for domain-based language processing (Magnini et al, 2002a; 2002b; Gliozzo, 2006). In this study, we use bootstrapping methods to automatically annotate word senses of Chinese …

Content truncated to 500 words