Visits And Word Class System Of Modern Chinese Content Words Grammatical Function Reconstruction

The deep analysis of the natural language, such as grammar checking, machine translation and text summarization, is more and more required to fit the rapid development of computer's automatic information-processing. The computer should be provided with the language information especially that of the grammar so as to ensure the correct analysis and automatic understanding. Generally speaking, the Chinese grammar system available is human-oriented, that is to say, it is good enough for human beings though, it is rather rough for the computer. But the realization of the automatic information-processing results from the deepening of the Chinese grammar researches and the specification of the grammar characterization, especially the reconstruction of the Chinese word class system suitable for computer natural language processing. Taking this as the goal, this dissertation made the following discussions.The first part, based on a corpus and taking statistic and introspective methods, makes a detailed survey of 3514 high frequency words' grammatical functions, which are nouns, verbs, adjectives and adverbs, and it establishes a database of grammatical function information. Besides, it classifies the 3514 high frequency words into 676 classes, in thorough accordance with the guideline of grammatical function. The result is that there are 364 classes with only one word, 107 classes with two words, and 205 classes with more than two words, among which, the first two classes are attributed to the lexical base, and the 205 classes are formed as a new substantive word class system.The second part compares the word class system made by other researchers and our new one in terms of the syntactic function complexity and classifying results. As for the function complexity, 10.1% are the same while 84.1% of the old one is more complicated than the new one. This shows that the old one exaggerates the multi-functional phenomena in Chinese word system. As for the classifying result, that one word class in the new system contains more than one word classes in the old system exists. This shows that the conclusion of one word with a particular word class can only be made on detailed survey and exploration instead of feature finding.