Chinese Collocation Acquisition Method Research

Natural language is an important tool for people to communicate with others, transmit information, share knowledge and express feelings. To better understand and apply language information, scholars have carried out researches on natural language early, and thus a subject that combined linguistics, computer science, and mathematics has emerged, which is called Natural Language Processing(NLP). The main tool of NLP is computer, the processing object is natural language, the purpose is to let machine learn and understand the grammar, sentence structure and associated rules of natural language, ultimately achieve that machines can process natural language intelligently. Collocation is one of the most common basic phenomena of NLP. Master and apply the correct collocations well can be very helpful for people to learn language quickly and express their ideas with accurate and concise words. Thus, collocation research has become an important aspect in the study of NLP. The research on collocation acquisition method and technology, the construction of related repositories are very important in the NLP research. If we can acquire accurate and comprehensive collocation resources and apply them into the practical field of NLP, such as machine translation, speech recognition, information retrieval, question answering system and so on, it will be no doubt that the intelligence and performance of related systems will be enhanced greatly.At present, the research on collocation acquisition method and technology has attracted considerable attention in NLP. For different researchers adopt different research methods and theoretical framework, the definition and acquisition method of collocation are also diverse. Among them, the most common methods are frame-based and semantic-based collocation acquisition. Frame-based collocation acquisition method emphasizes on the distance and structural characteristics which embodied in the language examples, this method is universal. However, due to lack of consideration of linguistic background knowledge, collocations that got from the frame-based collocation acquisition method are not very well in terms of semantic association. Otherwise, semantic-based collocation acquisition method tends to depend on the researchers’ own language intuition, related syntactic and semantic information, this method is not universal. Both method has its own advantages and disadvantages. In this paper, a collocation acquisition method that combined word semantic association and association strength is proposed.In this paper, existing collocation acquisition methods and realted technologies are clarified. By analyzing Chinese grammar, syntax structure, internal semantic relations and related rules, a on collocation from the two aspects of semantic and structure is made. A concept of the basic form of collocation in text statement analysis is introduced. Besides, a quintuple structural description for the basic form of collocation is offered. A method that how to get quintuple from text corpus, optimize and classify quintuple, finally obtain collocation resources is proposed as well. This paper adopted the content of People’s Daily(1998) in January as corpus, when this method is used to getting independence collocation and related collocation, the recall rate of independence collocation is 64.73%, the accuracy rate is 80.65%, the recall rate of related collocation is 64.79%, the accuracy rate is 85.19%. It shows that this method has some value for practical operation in Chinese collocation acquisition.