Research On Word Sense Disambiguation And Keyword Expansion In Question Answering System

Question Answering System is an advanced form of information retrieval system. The advantage is that it allow users to ask questions in natural language and return the exact answer, which is different from traditional search engine which returns a list of related links. With the study of natural language processing,question answering system has entered into a new stage.Question Analysis include word segmentation, POS tagging, word sense disambiguation and keyword expansion. Word Sense Disambiguation is the basic core problem in the natural language processing. It plays an important role in machine translation, information retrieval and other fields. In the information retrieval, word sense disambiguation provide a basis for subsequent searches. When search engine understands the specific needs of users, the accuracy rate of result could improved. This is the key point of questions analysis. With the development of information society, the Internet is filled with amounts of information. In the provision of resources for us, it also give us difficulty to find the resources. Because, users hardly write all the combinations of question, the result accuracy and the recall rate are low. Query expansion method can improve search result to some extent, it have been studied extensively.Specifically, the paper focuses on the following key research points: First of all, the paper discusses the commonly used Chinese semantic resources in the Chinese information processing, Word segmentation method, word similarity computation method, word sense disambiguation method, and analyzes the word similarity method and sentence similarity method which based on hownet. In the paper, sentence similarity algorithm which based on hownet are used to solve the word sense disambiguation. Second, it is important to improve the precision of traditiona retriecal system in order to meet users. In this paper, the method is used which is based on the initial query and the synonyms which are expanded by TongYiCi CiLin and hownet.