| With the development of Internet, net resources became increasingly rich.However, traditional search engine exist many disadvantages. For example, it returned the web pages not the exact answers, the precision is very low. Furthermore, it was too difficult to understand the purpose because traditional search engine which based on the keyword index did not deal with the semantic information. But the user could ask the nature language questions in the Intelligent Question Answering System (IQAS) which returned the answers directly after analyzing and processing the questions. So it was said IQAS system was the new generation search engine.Recently, Ontology was paid more and more attention and got many applications in the Artificial Intelligence area. In the closed domain QA system, Ontology knowledge base could express the inner relation and framework reasonably and reduced the redundancy, which was benefit to drawing out the answers based on semantics. After analyzing the current IQAS, the paper imported ontology technique into Intelligent Question Answering System, and built a domain ontology faced to《Data Structure》. The ontology was the information foundation of semantic understanding. Then it would implement IQAS based on Semantic Web.There were four key modules in the IQAS design model.Firstly, Domain Ontology Construction. We use protege3.1 to build a small domain ontology for《Data Structure》, and use Jena to analytical ontology then persistent to Mysql database. Secondly,Offline Semantic Reasoning. In order to guarantee real-time response speed of the system, the reasoning sections as offline processing and persistent to database.Thirdly, Linguistic Analysis. when inputting a natural language question, Linguistic Analysis took splitting word, word markup(respectively use je-analysis and Viterbi algorithm),we can get a group of word-character set,then through question pattern matching,last we can get a Query Linguistic Block.Fourthly,Semantic Similarity Service. QLB and Objective Search Block (OSB) Set through Questions Similarity Matching module, the accuracy of Questions Similarity is decided by Vocabulary Similarity. Vocabulary include general vocabulary and domain vocabulary, the former use the semantic similarity algorithm based on Hownet. In the latter, a semantic similarity algorithm was brought forward, which weighted the semantic similarity based on semantic distance and based on context which calculated the concept similarity of parent and child ones in the ontology,it greatly enhance the accuracy of the Domain Vocabulary Matching.The research of the paper showed that, IQAS could make full use of the semantic information of domain ontology, and it also solved the problem of lacking semantic understanding of the current search engine to some extend and had a high correctness ratio. |
