基于CRF模型的短文本信息流话题提取

摘要

A topic extraction method based on CRF model is presented in order to extract topic from Chinese short text information flow more effectively.According to the characteristics of short text information flow, the similarity of key words in short text information flow is defined. Global normalization of the feature information is processed through context information, and then the global optimal value is obtained.CRF method provides sig-nificant benefits if to compare the CRF method with decision tree method based on the real short text information set.%为更有效地在中文短文本信息流中进行话题提取，给出了一种基于CRF模型的话题提取方法。根据短文本信息流的特点，定义了短文本信息流中关键词语相似度。充分利用上下文信息对特征信息进行全局归一化的处理，进一步得到全局的最优值。在真实的短信文本信息集上将此方法与决策树方法进行比较，取得了较明显的优势。