Designing And Building A Chinese Knowledge Search System Based On Encyclopedic Knowledge

With the development of the Internet , finding information throw search engines has becomes a part of people's everyday life. The users only need to enter a query keyword, and then search engines return the best matched page links from the Web. But search engines can not return the direct answers. Users need to filter the links to find the needed information. As early as the beginning of artificial intelligence technology, people start thinking about how to make computer directly answer people's problems using natural language, this technique is known as automatic question answering system. As the technology and resource constrained that time, related research has been limited to specific areas. With the development of Web2.0 technology, more and more people share their information on the Internet, one of the most representative is the kind of online encyclopedia knowledge websites like Wikipedia. Wikipedia now has gathered vast amounts of knowledge, the appearance of the knowledge makes build a more general filed of intelligent automated answering system possible. In this respect, this thesis will attempt to design and implement a common knowledge search system based on encyclopedia knowledge. Users can do search by using natural language, and the system returns the direct answer from encyclopedia knowledge. The main contents include the following aspects:Firstly, this thesis designs a three-layer structure of knowledge search system solution by using ontology as a mid-layer between encyclopedia knowledge and knowledge search engines. A more intelligent knowledge search system needs a more convenient knowledge expression form understood by machines. This thesis studies how to use ontology representing knowledge and being a knowledge search system's data storage.Secondly, design an ontology construction method using multiple encyclopedia knowledge websites. There are some excellent constructed ontologies and basic approaches to building ontology based on a single encyclopedia web site, but few people interest in building ontology based on multiple encyclopedia web sites. This thesis constructs ontologies of several encyclopedia web sites firstly and then combines them into a single ontology.Thirdly, propose ontology-based multi-layer inverted index retrieval method. The multi-layer inverted index fully takes into account the level of ontology knowledge representation, the entity index is divided into two layers according to the categories of knowledge in the ontology. It makes the search process more directly by this way. Fourth, design and implement a complete knowledge search engine system. The entire system is divided into wiki sites collection subsystems, ontology construction subsystem, indexing subsystem, retrieval subsystem and user interface subsystem.