基于语义Web的多功能情报信息自适应检索技术

为了解决传统依据关键字、概念和属性值等检索技术没有考虑不同主体中情报信息间的语义关系，无法提供令用户满意的情报检索结果问题，通过语义Web方法研究了多功能情报信息自适应检索技术。设计基于语义Web的多功能情报信息自适应检索模型，检索模型包括人机交互层、知识处理层以及知识存储层。在知识存储层建立本体语义模型。在知识处理层，从检索的多功能情报信息关键词中推断和其语义有关的概念，依据概念有关词汇对原始查询进行扩展，对语义相似度进行计算。在人机交互层，将用户反馈情况看作判断检索结果优劣的衡量因素，依据相似关键词获取相似结果集，按照反馈因子完成对所有相似结果的排序，把排序结果发送给用户，实现多功能情报信息自适应检索。结果表明，所提方法能够同时保证召回率与精度；排序误差率低；检索结果和用户查询结果最相符，检索结果准确，可令用户满意。可见所提方法检索性能强。

Research on adaptive retrieval of multifunctional intelligence information based on semantic Web

In order to solve the problem that traditional retrieval techniques based on keywords, concepts and attribute values do not consider the semantic relationship between information in different subjects and can not provide satisfactory information retrieval results to users, a multi-functional information adaptive retrieval technique is studied by semantic Web method. A multi-functional information adaptive retrieval model based on semantic Web is designed. The retrieval model includes human-computer interaction layer, knowledge processing layer and knowledge storage layer. An ontology semantic model is established in the knowledge storage layer. In the knowledge processing layer, the semantic related concepts are inferred from the multi-functional information keywords retrieved, and the original query is expanded according to the related concepts, and the semantic similarity is calculated. In the human-computer interaction layer, user feedback is regarded as a measure factor to judge the quality of retrieval results. Similar result sets are obtained according to similar keywords, and all similar results are sorted according to feedback factors. The sorting results are sent to users to realize multi-functional information adaptive retrieval. The results show that the proposed method can guarantee the recall rate and precision at the same time, the sorting error rate is low, the retrieval results are most consistent with the user''s query results, and the retrieval results are accurate and satisfying to users. It is obvious that the retrieval performance of the proposed method is strong.