Dependency Treebank-based English Nouns Syntactic Research

Nowadays, the syntactic annotation on raw corpora becomes a dominant tendency in corpus linguistics. As the source of obtaining syntactic structure and foundation of estimating the syntactic parsing, treebanks, annotated corpus, have been attached great importance by the scholars from theoretical and computational linguists. The large amount of distributional information on syntactic function of part-of-speech can make enormous contribution on the theoretical linguistic research.On the base of PVP (Probabilistic Valency Pattern) theory, this research aims to clarify the common collocations and syntactic function of nouns, make a comparison with other previous researches via a quantitative analysis of English dependency treebank and respective interpretations of four research questions.The raw materials of over 6200 English sentences are extracted from the texts of New Concept English and Advance with English, and the dependency treebank contains part-of-speech tags and syntactic structure tags of the sentences, including tags of the governor, dependent of each word, and the corresponding dependency relations. The annotation of treebank was completed in two phases:the automatic parsing of Stanford parser and manual verification of the previous results. All these data were automatically generated into the Excel form, and the quantitative report of the syntactic functions was analyzed by Excel in the light of two conditions where English nouns act the governor and dependent distinctively. After the series of work done previously, the typical and atypical syntactic functions of English nouns would be concluded according to the frequency.The ultimate findings demonstrate that the typical syntactic functions of English nouns analyzed in this research generally match the arguments of other theoretical researches on English nouns. Moreover, this research emphasizes the distribution and performance of the principal and subordinate syntactic functions, supplementing some rare functions at the same time. Eventually, suggestions are given to replenish the universal English dependency representations and dependency types in this research.