依存分析和HMM相结合的信息抽取方法  

Method of text information extraction based on dependency parsing and HMM

作　　者：袁璐[1] 蒙祖强[1] 许珂[1] 

机构地区：[1]广西大学计算机与电子信息学院,南宁530004

出　　处：《计算机工程与应用》2012年第9期138-140,共3页Computer Engineering and Applications

基　　金：国家自然科学基金(No.61063032);广西教育厅科研基金项目(No.201012MS010)

摘　　要：信息抽取是文本信息处理的一个重要环节,当前的信息抽取研究工作大多针对半结构化的文本。针对自由文本,提出一种依存分析和HMM相结合的文本信息抽取算法,该算法在运用依存分析对句子进行浅层句法分析的基础上制定相应规则,形成输入序列,结合HMM易于建立、适应性好、抽取精度较高的优势,实现自由文本的信息抽取。实验结果表明,新的算法在召回率、准确率和正确率指标上均有良好的性能,说明了算法的有效性,为文本信息的抽取提供了新思路。Information extraction is an important part of text information processing. The current information extraction researches mostly focus on semi-structured text. It proposes a novel text information extraction algorithm based on the combination of dependency parsing and HMM. The algorithm formulates appropriate rules based on applying dependency parsing to shallow syntactic analysis of sentences, forming the input sequence of HMM to achieve free text information extraction combining the advantage of easily building, good adaptability and high extraction accuracy of HMM. Experimental results show that the new algorithm has very good performance on recall rate, accuracy and correct rate.

关 键 词：信息抽取 自由文本 隐马尔可夫模型 依存分析 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]