Auto-Discovery of NVEF Word-Pairs in Chinese

Auto-Discovery of NVEF Word-Pairs in Chinese Jia-Lin Tsai, Gladys Hsieh and Wen-Lian Hsu Institute of Information Science, Academia Sinica Nankang, Taipei, Taiwan, R.O.C. {tsaijl,gladys,hsu}@iis.sinica.edu.tw Abstract A meaningful noun-verb word-pair in a sentence is called a noun-verb event-frame (NVFE). Previously, we have developed an NVEF word-pair identifier to demonstrate that NVEF knowledge can be used effectively to resolve the Chinese word-sense disambiguation (WSD) problem (with 93.7% accuracy) and the Chinese syllable-to-word (STW) conversion problem (with 99.66% accuracy) on the NVEF related portion. In this paper, we propose a method for automatically acquiring a large scale NVEF knowledge without human intervention. The automatic discovery of NVEF knowledge includes four major processes: (1) segmentation check; (2) Initial Part-of-speech (POS) sequence generation; (3) NV knowledge generation and (4) automatic NVEF knowledge confirmation. Our experimental results show that the precision of the automatically acquired NVEF knowledge reaches 98.52% for the test sentences. In fact, it has automatically discovered more than three hundred thousand NVEF word-pairs from the 2001 United Daily News (2001 UDN) corpus. The acquired NVEF knowledge covers 48% NV-sentences in Academia Sinica Balanced Corpus (ASBC), where an NV-sentence is one including at least a noun and a verb. In the future, we will expand the size of NVEF knowledge to cover more than 75% of NV-sentences in ASBC. We will also apply the acquired NVEF knowledge to support other NLP researches, in particular, shallow parsing, syllable/speech understanding and text indexing. Keywords: noun-verb event frame (NVEF), machine learning, Hownet, WSD, STW 1 1. Introduction The most challenging problem in NLP is to program computers to understand natural lan- guages. For a human being, efficient syllable-to-word (STW) conversion and word sense disam- biguation (WSD) arise naturally while a sentence is understood. Therefore, in designing a natu- ral language understanding (NLD) system, two basic problems are to derive methods and knowledge for effectively performing the tasks of STW and WSD. For most languages, a sentence is a grammatical organization of words expressing a com- plete thought [Chu 1982, Fromkin et al. 1998]. Since a word is usually encoded with ploy-senses, to understand language, efficient word sense disambiguation (WSD) becomes a critical problem for any NLD system. According to a study in cognitive science [Choueka et al. 1983], people often disambiguate word sense using only a few other words in a given context (frequently only one additional word). Thus, the relationships between one word and others can be effectively used to resolve ambiguity. Furthermore, from [Small et al. 1988, Krovetz et al. 1992, Resnik et al. 2000], most ambiguities occur with nouns and verbs, and the object-event (i.e. noun-verb) distinction is a major ontological division for humans [Carey 1992]. Tsai et al. (2002) have shown that the knowledge of noun-verb event frame (NVEF) sense/word-pairs can be used effectively to achieve a WSD accuracy of 93.7% for the NVEF related portion in Chi- nese, which supports the above claim of [Choueka et al. 1983]. The most common relationships between nouns and verbs are subject-predicate (SP) and verb-object (VO) …

Content truncated to 500 words