| With the rapid development of Internet and WWW, the network information resources become more and more richer than before, search engines have already become the most popular tools to gain the network information. General search engines, such as BAIDU and GOOGLE, can search information from almost all kinds of fields. However, even those well-known general search engines also cannot fully satisfy people's needs when they search professional information in a specific field. In order to satisfy professional retrieval and dig out more information in some specific fields, vertical search engines arise at the historic moment. Vertical search engines can overcome the drawbacks of general search engines, but fully keyword-based vertical search engines still cannot make us get the ideal results in some situations, they can also perform poorly in many situations considering recall rate and effectiveness of relevance ranking.In this paper an ontology-based classification-index method for vertical search engine is proposed, semantic information is added into index by employing ontology-based text classification, and the recall rate and effectiveness of relevance ranking can be improved. Firstly, a classification system is designed on the basis of domain ontology, which implements fine-grained and multi-class text classification. Secondly, classification of information is written to the index. Based on the structure of Lucene index, the logical and physical structures of classification-index are designed so that category information and keywords are reasonably combined into an integrated classification-index. Finally, the algorithms of creating and searching classification-Index are designed; the methods of compressing and merging classification-Index are also discussed in this paper.The maritime ontology and web were taken as an example in this paper, and extensive experiments were made in the stage of index creation and retrieval. The experimental results indicated that the design of classification-Index. System performed high index creation speed and low coefficient of expansion; and retrieval based on the classification-index can gain higher recall and better effectiveness of relevance ranking. |
