商品评论的信息量化方法与计算

Quantification method and calculation of commodity review information

网购受到越来越多消费者的青睐,已有商品评论成为消费者进行商品购买决策的重要参考依据,而针对商品评论中所包含的信息量的研究相对较少。该方法以香农信息论为基础,首先从概率角度对商品评论文本进行信息量化;再根据商品评论中不同关键字词对消费者购买决策影响程度不同,建立加权信息量化模型;最后考虑不同评论文本之间存在语义重复、商品评论发布时间对消费者影响不同,通过对评论文本进行向量化表示、文本聚类最终得出互信息下的信息量化模型。对100条手机商品评论进行信息量化,得出在不同模型下同样的商品评论所包含的信息量不同,而综合考虑语义与时间维度的商品评论信息量与人的主观认知更为贴切。

Online shopping has been favored by more and more consumers. Existing commodity reviews have become an important reference for consumers to make purchasing decisions. However, there is relatively little research on the amount of information contained in commodity reviews. This article is based on Shannon＇ s information theory. Firstly, the information of the commodity review text is quantified from the perspective of probability; then according to different influences of different keyword words on the consumer purchase decision in the commodity review, the weighted information quantification model is obtained; finally, the semantic duplication and commodity review between different comment texts are considered. The release time has different influence on the consumers. By vectorizing the comment texts, the text clustering finally leads to the information quantification model under mutual information. In this paper, hundreds of commodity reviews are analyzed in experiments. It is concluded that the information amount of commodity reviews under different models is different, and the amount of commodity review information considering semantics and time dimension is more appropriate than human subjective cognition.