A Chinese Parsing Model Based on Corpus,Rules and Statistics

It is of great significance to take advantages of both rule-based approach and statisticsebased approach to improve the performance of parsers. For this purpose. the authors designed and implemented a new model for Chinese dependency parsing, called CRSP（Corpus, Rule and Statistics -based Parser）. The originality of CRSP is that dependency parsing is viewed as both statistics-based tagging process and a single statistics based tagging model is successively used for part of speech （POS） tagging and upper dependency relation （UDR） tagging. In this paper, we first give introduction to the design philosophy of CRSP, and then discuss the knowledge acquisition from an analyzed corpus to support the parsing process, and describe the statistic model for both POS tagging and UDR.tagging. The experiments show that the model is successful.