Research On Chinese Phrases Automatic Recognition Based On Frame Semantics

Chinese Phrase recognition is one of the research focuses in natural language processing all along. It plays a very important role in research of information extraction, QA system and machine translation. Phrase automatic recognition is a sub-question of syntax parsing. Furthmore, parsing is a fundemental task for frame semantic analysis. Phrase automatic identification applied to frame semantic analysis has particular requirements, so single phrase automantic recognition does not satisfy the requirements of frame semantic analysis.This thesis is based on the task of frame semantic analysis and employs various methods of machine learning to identify phrases in texts. We also construct different identifying systems and acquire satisfying result. This paper mainly focuses on the problems as following:1.Tranformation-based noun phrase recognition. Focusing on the aim of Frame semantic analysis, through analyzing practical corpus, this paper presents the definition of noun phrase for the frame semantic analysis. We introduct the transformation-based mothd and algorithm, then offer the transformation conditions in noun phrase recognition process. The results of experiments indicate that the transformation-based method is useful in recognition progress of noun phrase. Finally, though the results analyzing, we put forward the improvement method.2. Using maxmum entropy model to constructing the noun phrase recognizing model. In the process of recognition, firstly ,we should chose the features for the model, then erect the recoginition system which includes training model and annotating the texts. Through performing the system, it shows a better result.3.Verb phrase recognition based on rules. Through analyzing the corpus and aiming at the goal of recognition, we define the definition of verb phrases which suits for the Frame semantic analyaing. We also define a rule depending on the Functional unification grammar and use the rule to recognize the verb phrases. The results of the experiments show higher recall and precision.4.Chinese syntax and semantic parsing based on the frame semantic. Thsis thesis introduces the construction of Chinese FrameNet. Then we present a Chinese syntax parsing model and a Chinese semantic parsing model. Finally, we demostrate that these models are feasible through analylzing practical instances and will deeply research on the auto tagging with the Chinese Frame semantic.