Development And Application Of Domain Specific Semantic Retrieval System Of Institutional Repository

In the promotion of the open access campaign, institutional repository, as an important tool for the free spread of research products in network, play an important role in promoting the preservation, academic exchange and sharing of academic information. In recent years, many well-known domestic and foreign universities and academic institutions have established their own institutional repository. Researches have found that most institutional repositories are still using the retrieval technology based on keywords, but because the information resource of which they contain are abundant and highly specialized, such retrieval techniques often can not satisfy the users.In order to understand the user search intent better and improve the recall and precision of the IR’s retrieval system, this paper uses the word form standardzation, full-text retrieval, query expansion, word sense disambiguation and some other key technologies to design and implement a domain-specific semantic retrieval system of Institutional Repository, and proposes a semantic similarity algrithm based on WordNet. The workflow of the retrieval system consist of two stages. The first stage is document processing. The speech tagging and word form standardization sub-module unifies the word forms of the document. The stop word processing sub-module remove the stop words in the document. And the index sub-module create indexes for the documents. The second stage is user input processing. First, unifies the word form and remove the stop words as the same stage.Word sense disambiguation and expansion sub-module determines the meaning of the query word and extends the user input. Domain filter sub-module filter the pre-extend words, and remain the words of this domain. Finally, the search sub-module generate search queries and returns the results to the user.Through experimental test, the semantic retrieval system has better precision and recall rates compared to the retrieval system based on key words.