Semantic Representation and Composition for Unknown Compounds in E-HowNet

Semantic Representation and Composition for Unknown Compounds in E-HowNet Yueh-Yin Shih, Shu-Ling Huang, and Keh-Jiann Chen Institute of Information Science, Academia Sinica, Taipei, Taiwan {yuehyin,josieh}@hp.iis.sinica.edu.tw, kchen@iis.sinica.edu.tw Abstract. This paper describes a universal concept representational mechanism called E-HowNet, to handle difficulties caused by unknown words in natural language processing. Semantic structures and sense disambiguation of unknown words are discovered by analogy. We intend to achieve that any concept can be defined by E-HowNet and the representation is near-canonical. The design for easy semantic composition and decomposition makes the automation of semantic processing for unknown words, phrases and even sentences possible. Keywords: unknown word, compound, E-HowNet, semantic composition, analogy, semantic similarity. 1 Introduction The occurrences of unknown words cause difficulties in natural language processing. The word set of a natural language is open-ended. Because novel words are created daily for expressing new concepts and new inventions, it is impossible to collect them completely in a dictionary. Morphology of Chinese allows new words to be generated by compounding and affixation. Orthographically, such compounding and affixation are represented by combinations of characters. In general, each Chinese character represents a morpheme. Each Chinese morpheme carries meanings and most are polysemous. New words are easily constructed by combining morphemes and their meanings are the semantic composition of morpheme components. (Of course there are also exceptions of semantically non-compositionally compounds.) It is impractical to construct a lexical database containing semantic information for all words. Therefore, it would be valuable to have a good design of semantic representation system which can easily compose the necessary information for new words. A compound is a combination of two or more words/morphemes bounded together by morpho- syntactic relations such as modifier-head, verb-object and so on. According to an inspection on the Sinica Corpus [1], 3.51% of the word tokens in the corpus are unknown, which contains about 80,000 entries. Among them, about 51% of the word types are compound nouns, 34% are compound verbs and 15% are proper names. In this paper, we introduce our concept representation mechanism, called E- HowNet and demonstrate the capabilities of this system in representing senses for unknown compound nouns, due to its advantage of easy semantic composition and decomposition. This paper is organized as follows. Section 2 describes the characteristics of E-HowNet. In section 3, unknown compound nouns are taken as examples to demonstrate the analogy approach in discovering the semantic structures of unknown words in E-HowNet. Conclusion and future work are drawn in section 4. 2 E-HowNet Lexical knowledge representation has become a major research area for natural language processing in recent years. To bridge gaps between natural language representations and conceptual representations, 378 we had proposed a frame-based entity-relation knowledge representation model called E-HowNet, which was evolved from HowNet [2] to encode concepts. It extends the word sense definition mechanism of HowNet and uses WordNet synsets [3] as vocabulary to describe concepts. In principle, the qualia structure is the major features for a nominal-type concept [4] and event frames are for eventive concepts …

Content truncated to 500 words