基于规则推理的语义检索若干关键技术研究  

Research on Critical Technologies of Semantic Retrieval Based on Rule Reasoning

作　　者：马森[1,2] 赵文[2,3] 袁崇义[1,3] 张世琨[2,3] 王立福[2,3] 

机构地区：[1]北京大学信息科学技术学院,北京100871 [2]北京大学软件工程国家工程研究中心,北京100871 [3]北京大学信息科学技术学院软件研究所高可信软件技术教育部重点实验室,北京100871

出　　处：《电子学报》2013年第5期977-981,共5页Acta Electronica Sinica

基　　金：国家973重点基础研究发展计划(No.2009CB320706);解放军总装备部预研项目(No.513150202)

摘　　要：针对专业领域复杂的检索需求,目前相关研究采用基于语义的方法来扩展检索范围并提高准确度.在语义推理方面,目前搜索引擎通常直接采用语义网中的推理算法,推理效率不高.在排序方面,基于关键字的搜索引擎的排序算法也不适合对语义检索结果进行排序.针对上述问题,本文给出了基于语义网的语义规则建立方法,并提出了一种基于闭合世界假设的反向链接推理算法,提高推理效率,同时给出了一种基于特征相似性排序算法,使检索结果排序方式更加符合语义检索的特点.基于本文提出的方法,构造了语义搜索引擎MaterialHub,实验表明该搜索引擎提高了检索的准确率和查全率,有较好的查询响应时间,并已经得到实际应用.For some complex specified domain retrieval requirements,many researches use semantic related technologies to resolve such problems.In terms of rule inference,it always leverages the inference algorithm of Semantic Web directly.However,the efficiency is not good.On the respects of searching results ordering,the ordering algorithm for search engine,which is based on keywords,is not suitable for ranking on searchresults generated by semantic retrieving.Focusing on the above issues,this paper proposes a semantic rule modeling method,and gives a new rule reasoning algorithm based on closed world assumption backwards reasoning chain to get higher inference efficiency compared to most semantic inference engines.Moreover,this paper proposes a new ordering algorithm based on feature similarity.Taking advantage of the above methods this paper describes,the search engine-Material Hub has been built up.Experiments show this semantic search engine improves the searching precision rate,recall rate,and rational responding time.So far,this system has been applied in industry.

关 键 词：语义检索 语义规则 语义规则推理 语义相似性排序 

分 类 号：TP306[自动化与计算机技术—计算机系统结构]