Using Corpus-based Linguistic Approaches in Sense Prediction Study

Using Corpus-based Linguistic Approaches in Sense Prediction Study∗ Jia-Fei Hong a, Sue-Jin Ker b, Chu-Ren Huang c and Kathleen Ahrens d aInstitute of Linguistics, Academia Sinica, Taiwan, No. 128, Section 2, Academia Road 115, Taipei, Taiwan jiafei@gate.sinica.edu.tw bDepartment of Computer Science and Information Management, Soochow University, Taiwan, No.56, Section 1, Guiyang Street 100, Taipei, Taiwan ksj@cis.scu.edu.tw cFaculty of Humanities, The Hong Kong Polytechnic University, Hong Kong The Hong Kong Polytechnic University, Hong Hum, Hong Kong churenhuang@gmail.com dLanguage Centre, Hong Kong Baptist University, Hong Kong Hong Kong Baptist University, Kowloon Tong, Kowloon, Hong Kong kathleenahrens@yahoo.com Abstract. In this study, we propose to use two corpus-based linguistic approaches for a sense prediction study. We will concentrate on the character similarity clustering approach and concept similarity clustering approach to predict the senses of non-assigned words by using corpora and tools, such as Chinese Gigaword Corpus, and HowNet. In this study, we would then like to evaluate their predictions via the sense divisions of Chinese Wordnet and Xiandai Hanyu Cidian. Using these corpora, we will determine the clusters of our four target words ---- chi1 “eat”, wan2 “play”, huan4 “change” and shao1 “burn” in order to predict their all possible senses and evaluate them. This requirement will demonstrate the visibility of the corpus-based approaches. Keywords: Lexical ambiguity, sense prediction, corpus-based approach, character similarity clustering, concept similarity clustering, evaluation. ∗ Copyright 2010 by Jia-Fei Hong, Sue-Jin Ker, Chu-Ren Huang and Kathleen Ahrens 1 Introduction Our goal in this study of sense prediction is to generate solutions for lexical ambiguity in general. In particular, we will look at words without lexically-assigned senses and try to predict the range of senses each word form may have. Since lexical information of senses of these words is not available, we propose to use corpus-driven distribution as the main information is prediction. We will determine the collocation clusters of our target word through characters, semantic features, and concepts by using corpora and tools, such as Chinese Gigaword Corpus, Chinese Wordnet and Xiandai Hanyu. Our study showed the feasibility of sense-prediction without lexically assigned senses. If a word has more than two senses at the same time, then it is usually called a lexically ambiguous word, for example, bank. We need to divide its correct senses and assign its appropriate senses to the different contexts. However, “lexical ambiguity” is presented in several different related researches to present and refer the same target. In WordNet, the definition of a lexically ambiguous word is the ambiguity of an individual word or phrase that can be used (in different contexts) to express two or more different meanings. WordNet researchers also regard polysemy and lexical ambiguity as synonym. In the case of the PACLIC 24 Proceedings 399 previously related lexical ambiguity, several studies have concentrated on the corpus-based and computational perspective included: Peng et al. (2007), Xue et al. (2006), Chen et al. (2005), Moldovan and Novischi (2004) …and so on and they took several different approaches: used the corpus-based approach, an adaptive system, divided …

Content truncated to 500 words