融合用户需求和商品特点的评论标签生成算法研究

Research on the Generation Algorithm of Comment Labels Integrating User Needs and Product Features

如今网络购物用户规模不断上升,评论数据成为消费者进行购买决策的主要依据。同时,商家需要利用评论明确用户。消费者没有充足时间、合适语言进行商品评论成为一个亟待解决的问题。然而,仅有少数网购平台在评论页面提供标签,只能说明商品局部特点,没有体现用户个性化需求。论文通过分析用户评论中彰显用户个性化的语言描述特点,不同种类商品的描述方面差异,设计能够自动生成个性化评论标签的算法,用标签充分体现商品特点与用户需求。将单个用户作为研究对象,其全部评论作数据集,在TextRank模型中加入词性-句法特征分析,捕获用户评论语言描述风格的内在联系。对比单独使用TextRank模型等三个成熟模型生成标签,算法效果更好。

Nowadays,the scale of online shopping users continues to increase,and review data has become the main basis for consumers to make purchase decisions.At the same time,businesses need to use comments to identify users.Consumers do not have enough time and appropriate language to comment on products has become an urgent problem to be solved.However,only a few on⁃line shopping platforms provide tags on the review page,which can only illustrate the local characteristics of the product and fail to reflect the personalized needs of users.The paper analyzes the characteristics of the user's personalized language description in user reviews and the differences in the description of different types of products.It designs an algorithm that can automatically generate personalized review tags,and uses tags to fully reflect the characteristics of products and user needs.Taking a single user as the re⁃search object and all their comments as a data set,part-of-speech-syntactic feature analysis is added to the TextRank model to cap⁃ture the internal connection of the user's comment language description style.Compared with using three mature models such as Tex⁃tRank model to generate labels,the algorithm effect is better.