作　　者： ; ; ; ;

机构地区： 湖南大学信息科学与工程学院软件学院

出　　处： 《系统仿真学报》 2007年第21期4926-4931,共6页

摘　　要： 应用隐马尔可夫模型是文本信息抽取的一种重要方法。对于网上不同来源的文本,由于其格式很不相同,进行混合训练,一般难以得到较优化的模型。将聚簇应用到文本信息抽取中,首先通过一种改进的k-平均方法对训练文本的Markov链模型进行聚簇,然后训练各簇的隐马尔可夫模型,提出了一种基于聚簇隐马尔可夫模型的文本信息抽取算法(C-HMM)。对700篇网上不同来源的文本进行信息抽取仿真实验,结果表明,新的算法能有效地提高抽取性能。 Using Hidden Markov model is an important approach for text information extraction, The form is dissimilar for texts which are from different resource of network. The optimal model is commonly difficult to obtain by hybrid training texts Clustering was applied to text information extraction. Clustering was given to Markov Chains of training texts through an improved approach of K-mean, and Hidden Markov model was trained out through every cluster. An algorithm of text information extraction based on clustering hidden Markov model （C-HMM） was proposed, A simulation experiment of information extraction was tried on 700 texts from different resource of network. Results show that the performance of extraction can be improved effectively by using the new algorithm.

关 键 词： 聚簇 马尔可夫链 隐马尔可夫模型 信息抽取

领　　域： [自动化与计算机技术] [自动化与计算机技术]