The Research And Implementation On Wordnet-based Sentence Similarity Of Automatic Question Answering System

With the rapid development of information technology, there are more and moreinformation and data on the Internet. The demand of information retrieval is increased. Howto search effectively the needed information from a mass of information is a difficulty andchallenge. Using keywords, it is mainly used for information retrieval by the search engine.There are a lot of defects of the traditional search engine. The search process is complex, andit returns too much information. Therefore, scholars have proposed a concept of a newapplication: Automatic Question Answering System.Automatic Question Answering System accepts the questions by the user using naturallanguage. It returns an accurate, concise answer to the user. The most common AutomaticQuestion Answering System is based on frequently asked questions (FAQ) Knowledge Base.The sentence similarity calculation is the core of a FAQ Automatic Question AnsweringSystem. Vector space model TF-IDF algorithm and based on semantic sentence similarityalgorithm are common sentence similarity algorithm. The new sentence similarity algorithmof this thesis is based on WordNet and Edit Distance method. WordNet is a large lexicaldatabase of English. It is organized by the Synsets and the semantic relation between them.WordNet provides great helps natural language processing and information retrieval.Firstly, the Automatic Question Answering System technologies are introduced. Then,how to use the WordNet to calculate sentence similarity will be described. Based on thecommon sentence similarity computing methods, a new sentence similarity computingmethod is proposed. This new method combines two aspects. This method calculates thesentence semantic similarity with WordNet by Maximum Matching Problem method. And italso considers word order of the sentence similarity by Edit Distance method. A FAQAutomatic Question Answering System application is designed and implemented using thenew sentence similarity computing method. The test will be executed to determine the weightof the new method based on the system. In addition, compared with the different sentencesimilarity methods, the experiment will verify the validity of the new method. Finally, theconclusion and the problems of my works, and the prospects will be given.