基于多标记学习的汽车评论文本多性能识别MultiplePerformances.PDFVIP

文章编号：1003-0077 （2015）00-0000-00 基于多标记学习的汽车评论文本多性能识别 1 1,2 1,2 张 晶 李德玉 王素格 （1.山西大学 计算机与信息技术学院，山西 太原 030006 ； 2. 山西大学 计算智能与中文信息处理教育部重点实验室，山西 太原 030006 ） 摘要：针对汽车产品评论文本中出现的多方面性能，提出一种基于多标记学习的汽车评论文本多方面性能 识别方法。首先，结合文本挖掘方法，利用多标记文本特征选择方法选取特征，将非结构化的文本转化为 结构化的多标记数据集。在此基础上，使用4 种多标记分类方法，对待识别的评论文档标注一个或多个方 面标记。最后，以 8 种多标记评价指标评估方面识别的性能。在新浪汽车评论语料上的实验表明，方面识 别的子集准确率达到了95%，验证了方法的可行性。 关键词：多标记学习；文本处理；汽车评论；多方面识别 中图分类号：TP391 文献标识码：A Multiple Performances Identification for Car Review Texts Based on Multi-label Learning Zhang Jing1 Li De-yu1,2 Wang Su-ge1,2 (1.School of Computer and Information Technology, Shanxi University, Taiyuan, Shanxi 030006, China; 2.Key Laboratory of Computational Intelligence and Chinese Information Processing of Ministry of Education, Shanxi University, Taiyuan, Shanxi 030006, China) Abstract: Aiming at the characteristics of the multi-aspect performance appeared in the automotive product reviews, this paper proposed a novel method for recognizing the multiple aspects of performance about car comment text based on multi-label learning. Firstly, appropriate words were selected as features by multi-label text feature selection method combined with the text mining technology, and then, the unstructured document corpus are transformed into structured multi-label dataset. After that, we finished marking one or more aspect tags for the unrecognized comment text with four multi-label classification methods. Finally, the recognition accuracy of multiple aspects was assessed by eight multi-label evaluation metrics. On the sina car review corpus, experimental results indicate the subset