基于HMM/BP混合模型的文本信息抽取研究

Text Information Extraction Research Based on HMM and BP Network Hybrid Model

作为自然语言处理的一个分支,文本信息抽取成为了提取大量文本信息中有用信息的重要手段。介绍了目前在信息抽取领域中应用广泛的两种技术方法:HMM和BP网络模型,分析了各自的优缺点,并在此基础上提出了一种基于两者的混合模型,该混合模型通过BP网络优秀的分类甄别能力来弥补HMM在分类方面的不足,而通过HMM强大的时域建模能力来弥补BP网络建模能力弱的问题,因此该模型具有强大的建模能力、分类性以及适应性强等特点。实验证明,相比传统的HMM以及BP网络模型,该混和模型在精确度和召回率上有了10%～15%的提高。

As a branch of natural language processing, the extraction of useful information in large text , the text information extraction became an important means. Introduce the information extraction widely used two kinds of technical methods： HMM and BP network model, analyze their advantages and disadvantages and on this basis propose a hybrid model, based on two models mentioned above. In this model, the classification by BP network capacity is to make up for deficiencies in the classificationof HMM, HMM through strong time-domain modeling capabilities to make up for weak BP network modeling problem,so the hybrid model has strong modeling capabil- ities, classified and adaptability, etc. Experimental results show that compared to the traditional HMM and the BP network model, hybrid model in precision and recall rate is on the increase by 10% - 15%.