Research On Hierarchical Chinese Functional Constituent Identification

syntactic parsing is always an important task in the area of natural language processing, the main goal is to construct a model which is based on grammatical knowledge and statistical information to automatically build a hierarchy tree which consists of the sentence unit and the relation of each unit. As far as we are concerned, the main approaches for this task lies on two aspects: constituent parsing and dependency parsing, both methods can produce the syntactic structure based on grammatical rule and condition. However, after these two methods have been presented, there is no extra research on other constituents, such as the function of the words on the syntactic tree, it still remain on the original structure. But such constituents are of importance on many NLP tasks, for example in machine translation, we can use these information during the process of word alignment, in dependency parsing, we can use such information as filter ru1 e during beam search. Under such condition, we raise the research on hierarchical Chinese functional constituent identification.Currently, there exists only a small amount of the related research on Chinese functional constituent identification, furthermore, there is also not much on the hierarchical research on such topic. Under such condition, we raise the method to construct hierarchical Chinese functional constituent corpus, and present the analysis method on different level of hierarchical Chinese functional constituent tree.In this paper, our work concentrates on two aspects:(1)Firstly, following the similar strategy of previous research on related topic, we bring forward a rule-based approach to extract corpus from Chinese Tree Bank to construct the hierarchical Chinese functional constituent tree, then, we extract the data from three levels which are clause level, base level and complex noun phrase level, the data we extract from the functional constituent tree are used for the task of hierarchical Chinese functional constituent identification.(2)Secondly, in the hierarchical Chinese functional constituent identification, we present a step-by-step method, which can extract the constituent from top level to the bottom, we also introduce a novel algorithm which make use of other result of other level as new features, the experiment results shows that this method can help to improve the performance of the baseline model.Based on the above analysis, we can find that the research on hierarchical Chinese functional constituent identification is of great value to process,we come up with a potential method to achieve this goal,the experiment results show that this research has a valuable research prospects.