A lexical semantic study of four-character Sino-Japanese compounds and its application to machine translation

Four-Character Sino-Japanese compounds are a productive word formation process in Japanese. There are many morpho-syntactic analyses on these compounds. However, little has been done on their lexical semantic structure. In this thesis I will provide a syntactically motivated classification system for these compounds, and a lexical semantic analysis of their structure. The lexical semantic analysis is extended to a potential application in Japanese-English Machine Translation. A lexical semantic analysis reveals that for compounds with a deverbal head, there is an argument relation between the constituents if the head’s lexical semantic requirement is fulfilled by the non-head constituent, while the relation is adjunct if it is not fulfilled. The constituents of compounds with a regular noun head are in an attributive relation, and the relation cannot seem to be determined by Lexical Semantics. Compounds with a de-adjectival head require more examples to draw firm conclusions because these compounds are rare.