부산대학교 도서관

As a generalized approach in natural language processing, pattern matching is seldom applied in syntactic parsing nowadays. In some applications on short text analysis such as microblog opinion mining, the sentences are characterized by obvious patterns. Thus automatic parsing by pattern matching may be more effective than general syntactic parsing method. This paper puts forward a lightweight algorithm of Matching Pattern (MP) acquisition to achieve the syntactic parsing on some specific Chinese text composed of short clauses. The key points of the algorithm are MP generation based on word/POS sequence and MP selection based on weight ranking of mapping between MP groups and sentence groups. Experiments show that this method performs well on Chinese corpus with the following features: (1) The sentences are mainly composed of short clauses, (2) Most of the clauses can be represented by limited patterns.