论文部分内容阅读
随着社交媒体的发展,不断增加的在线产品评论正在极大地影响电子商务市场,使得评论挖掘成为商业界与学术界共同的热点话题。针对中文产品评论的特点,本文提出一种基于领域本体的建模方法,通过建立评论挖掘模型来对产品评论的基本评价单元——“特征观点对”进行识别。该建模过程以设计科学研究方法论为指导。首先在模型设计阶段,构建面向产品评论的领域本体;然后在模型实施阶段,提出基于本体的特征观点对识别方法;最后在模型评价阶段,通过实验对评论挖掘结果进行评价。实验结果表明,本文提出的方法与其他基于统计的方法以及基于语义的方法相比,在性能上有明显提高,对克服口语化严重和语法不规范等问题具有良好的效果。此外,通过特征观点对的识别与统计,使产品评论这种非结构化文本转化为机器可读的、能理解的结构化表达并得到具有一定商业应用价值的信息。
With the development of social media, the increasing number of online product reviews is greatly affecting the e-commerce market, making review mining a hot topic of common interest both in business and academia. In view of the characteristics of Chinese product reviews, this paper proposes a domain based ontology modeling method, through the establishment of a review mining model to the product evaluation of the basic evaluation unit - “feature point of view” to identify. The modeling process is guided by the methodology of designing scientific research. Firstly, the domain ontology oriented to product reviews is constructed in the model design stage. Then, based on the ontology feature point of view, a recognition method is proposed in the model implementation stage. Finally, during the model evaluation stage, the evaluation results of the comment mining are evaluated. Experimental results show that compared with other statistical-based methods and semantic-based methods, the proposed method has obvious improvement in performance and has good effect on overcoming the problems of serious colloquialization and non-standard grammar. In addition, through the recognition and statistics of feature point pairs, the unstructured texts of product reviews are transformed into machine readable and comprehensible structured expressions and obtained information with a certain commercial application value.