【作者】 王宇；

【导师】 沈琦；

【作者基本信息】 北京工业大学 ， 软件工程（专业学位）， 2015， 硕士

【摘要】 煤矿安全事故在我国时有发生,网络中存在着大量的相关报道、政策法规、处理措施等信息,信息冗余、繁杂,用户在检索相关信息过程中经常得不到需要的结果。本文在研究煤矿安全领域特点的基础上,利用事件本体技术将该领域信息进行抽象化,作为领域知识来表示,并提出了一种基于事件本体的查询扩展方法,实现了语义检索,有效改善了传统搜索引擎基于关键字匹配原理造成的查全率和查准率过低的问题。首先,分析煤矿安全领域特点,发现以“概念”作为本体基本单元的传统本体构建模式缺乏动态性,并不能准确描述该领域知识,因此,提出了事件本体概念,并阐述了相关技术。其次,定义了事件类,事件类属性,事件类间关系,提出了事件类间相关性因子概念及其计算方法,并将其作为衡量事件类间关联程度强弱的依据,在此基础上提出了事件本体模型,为之后煤矿安全事件本体的构建奠定了理论基础。再次,提出了一种以领域文档作为数据源的本体学习技术,其中包括文档中事件类的抽取方法,类比网页链接算法提出的事件类重要度的改进算法,并通过实验验证了其效果。依据上述本体学习技术,设计并实现了重要事件类推荐系统,经过实验验证了该本体学习技术的可行性。之后,选择owl表示语言,利用Protégé编辑工具构建了煤矿安全领域事件本体。最后,提出了一种基于事件本体的查询扩展技术。利用分词和依存句法等自然语言处理技术分析查询项构成,对其中事件项进行事件类间及事件属性两方面的扩展,之后运用向量空间模型向量化查询项和领域文档,并进行相似度计算,排序输出结果文档,最终经实验验证了该查询扩展方法的优越性。更多还原

【Abstract】 Coal mining accidents in our country occur currently, there is a large number of relevant reports, policies and measures on the Internet, which is redundant and complicated. Users often can’t get the results that they want in the retrieval process. Based on the characteristics of coal mine safety field, the paper uses event ontology technology to make domain information abstraction as domain knowledge, and puts forward query expansion method based event ontology, that implements the semantic retrieval and effectively improves the traditional search engine’s problem of recall and low precision, because it’s based on keyword matching principle.First, the paper analyses characteristics of coal mine safety field and finds that putting "concept" as a basic unit of ontology construction can’t accurately describe the domain knowledge because it’s lack of dynamic. Therefore, it is suggested that the event ontology concepts, and explains relevant technology.Secondly, the paper expounded the concept of event class, the relationships between event classes and event class properties, then proposed the concept of correlation factor between event class and its calculation method, which is used as a measure of association degree between event class. At last, event ontology model is proposed, which is laid the theoretical basis of constructing event ontology in coal mining safety field.Again, the paper put forward an ontology learning technology that treated field document as data source, including extracting event class in the document, a calculation method for the important degree of event class based on page link algorithm that is effective through an experiment. On the basis of above algorithm, we designed and implemented a important event class recommendation system. An experiment verified the feasibility of the ontology learning technique. Later, we select the owl ontology representation language and use the Protégé ontology editing tool to construct event ontology in the field of coal mining safety.Finally, we proposed a query extension method based on event ontology. First, we use word segmentation and dependency syntactic technology to analyze the query terms and extend event term in event and event’s properties, then use vector space model to vectoring the query terms and document and calculate similarity,get the output results.The query expansion method is feasible through an experiment.更多还原