Dynamic QA Systems Using Knowledgebase

Nowadays, with the quick development of Internet, the amount of available information still is growing exponentially. This way the Internet has become an important resource to answer user‟s questions. The traditional search engine can help people search information conveniently to some extent, but it also may return so many relevant results mixed with a large number of irrelevant results. In this way, users can hardly get their desirable answers immediately. Consequently, people are in need of a new more intelligent information retrieval system which can satisfy their needs. Question-Answering (QA for short) system is such a new way of information retrieval. QA system provides us a with human-machine interface that support natural language. For one thing this mode conforms to the way how people ask questions, for another it has clear advantages over the traditional search engine on understanding the intent of a question. In addition, QA system in several functions more accurately than search engine because it returns the more accuracy answer. This thesis presents an on-going research of the architecture and operation of a dynamic Question Answering system. The main results of this thesis are the architecture to handle the smart answering. This system is designed for Natural Language Processing in open domain based on Hownet knowledgebase, based on the knowledgebase, Natural Language Processing can get better effect; the system is available for learning ability. The ability is realized by the function of database updating. This system aims at high speed and accuracy on searching answers. Three components were improved by improving their algorithm. In texts clustering, we proposed a new ant-tree algorithm to replace the traditional K-means algorithm. It speeds up the procedure of clustering. During the key words extension, we combine the classical algorithm with HowNet knowledgebase in order to keep the high accuracy of understanding natural language. In the answer extraction, we improved the obsolete algorithm by considering necessary factors. Through testing a prototype of dynamic QA system, the improved algorithm can be proofed to be effective. In addition, the new architecture combined with the improved algorithms is the main contribution of this thesis.