Computer-oriented "single V_ + A_ Single Complement Structure Study

Since 1990s, as the range of verb study was widened gradually to verb phrases, predicate-complement structure, particularly this sort of structures of " Verb + Attached complement " attracted wide attention and became the important studying topic, due to its characteristic and its difficulty of learning. But we can see rare linguistic study of machine translation, information retrieval and information extraction among the large number of studying achievements.For meeting the demands raised by the informational society to the study of Chinese main body study, the author took this important two-character structure of " Uni-morpheme verb + Uni-morpheme adjective" predicate-complement structure as the studying target. Focusing on the aim of serving for Chinese informational procession, the author conducted some new researches and suggested the using of the combination of corpus method and introspection method, so that the accuracy of recognition and generation of this structure might be realized.This item of study was performed in two directions. Firstly, 17866 connecting pairs of the " Uni-morpheme verb + Uni-morpheme adjective " structure were extracted from a tagged corpus of 100 million Chinese characters. Tagging out the types of connecting one by one, the author investigated the beginning point of parsing by computer, how difficult it might be and what inspirations we might obtain from the corpus approach and its limitations. Secondly, after the comparison of several kinds ofdictionaries, the author picked out the two sets of words of " Uni-morpheme verb " and " Uni-morpheme adjective " which could be used in the predicate-complement structure. In these two sets of words, every element is a sense of word. By means of introspection, the author widened the range of instances used in the predicate-complement " Uni-morpheme verb + Uni-morpheme adjective " structure, and investigated the semantic selection rules between the two constituents and the relative nouns. These rules contain the aspects of position, number and direction. The aim is to investigate the effects of the semantic researches in aspect of automatic parsing. Meanwhile the author compared the corpus approach and the introspection approach. After that, the author distinguished the 5 types of " Uni-morpheme verb + Uni-morpheme adjective" predicate-complement structure, predicate-objective structure and " Uni-morpheme verb + Uni-morpheme adjective + Noun" structure, and aroused the disambiguity in syntax and semantics. Finally, beginning from the meaning-referring of " Uni-morpheme verb", " Uni-morpheme adjective", the author summed up the relationship between the 8 semantic patterns which could be expressed by "Uni-morpheme verb +Uni-morpheme adjective" predicate-complement structure and 37 types of surface sentence pattern.