Researches On PCFG-Based Parsing Method For Chinese Language

Parsing is a fundamental problem in natural language processing. Many natural language processing tasks, such as machine translation, information acquisition, automatic abstracting, have to depend on the precise parsing results to be finally resolved satisfactorily. At the same time, parsing techniques also help to solve problems similar with parsing in other areas. On the other hand, Parsing is also a basic problem in dealing with languages which are the carrier of the human mind, thus, research on parsing will be helpful for us to find the nature of human intelligence. Therefore, natural language parsing research has important theoretical value and profound philosophical meaning.In the Statistics-based parsing methods, the two most critical issues are the disambiguation model and parsing algorithms, they determine the accuracy and efficiency of parsing. This paper works from both sides, proceeds a PCFG-based parsing method combining context information, the main research work are as follows:1. Study of existing commonly used statistical model of parsing and parsing algorithm, and comprehensively analyze and compare their performance;2. Based upon above studies, we propose a disambiguation model including some context information;3. Proceeding our own parsing algorithm by expanding GLR algorithm;4. Demonstrating the effectiveness of our method by experiments.We adopt the People's Daily Corpus in January, 1998 and the machine translation Treebank produced by Institute of computing technology, the Chinese Academy of Sciences to be the training corpus, and summarized a series of grammar rules from the corpus for experiments. The small-scaled experiments we implemented proved that our parsing method is relatively efficient and accurate.