用规则抽取句子中事件信息  

Using Rules to Extract Event Information from Sentences

作　　者：孙荣[1,2] 周文[1] 刘宗田[1] 

机构地区：[1]上海大学计算机工程与科学学院,上海200072 [2]扬州市房地产信息中心,江苏扬州225003

出　　处：《小型微型计算机系统》2011年第11期2309-2314,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(60975033)资助;上海市重点学科建设项目(J50103)资助

摘　　要：信息抽取是数据挖掘的重要课题.目前的研究主要通过机器学习的方法对信息进行抽取.但是机器学习对训练数据的质量要求高,学习过程中参数设置复杂.而利用事先构建好的规则可以简单有效的从文本中提取事件信息.提出一种基于抽取规则对句子中的事件信息进行抽取的方法,摆脱了繁杂的机器学习过程.该方法利用本体对动词与事件角色匹配规则、事件角色抽取规则、时间信息抽取规则和地点信息抽取规则进行定义,用OWL对这些抽取规则进行了描述,然后应用这些规则抽取句子中的动词词义信息、事件角色信息、时间信息和地点信息,并用本文提出的一种新评测指标对事件信息进行评测.实验表明该方法从句子中抽取事件信息是有效的.Information extraction is an important task of data mining. Recent researches extract information by techniques of machine learning, which needs high-quality training data and complex parameter setting. However, it is simple and effective to extract event information by the rules constructed in advance from texts. This paper presents an approach to extract event information from sentences based on extraction rules and avoids the Complex process of machine learning. The method uses ontology to define the extraction rules of verb matching with event role, event role, time and location respectively and describes them based on OWL （ Ontology Web Language）. Then these rules are utilized to extract verb sense information, event role information, time information and location information. A novel evaluation of event information is introduced to evaluate the information extracted from text. The experiment results show that the approach is effective.

关 键 词：规则 本体 句子 信息抽取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]