面向用户评论的店面画像构建研究——以美团网为例

Research on the Construction of Storefront Portrait Oriented to User Comments:Taking of Meituan as an Example

[目的/意义]从美团餐饮用户评论数据中构建店面画像评价维度和店面画像得分,助力门店提升数字化和智能化管理水平,实现用户价值深度挖掘及门店最大化收益。[方法/过程]首先抽取关键词进行BERT词向量表示,通过高斯混合维度聚类、维度识别及维度词挖掘来构建店面画像评价维度,然后构建领域情感词典和句法分析获得维度词得分,最后为构建的店面画像进行打分。[局限]不足之处在于本次实验仅以武汉地区的美食自助餐评论为主,其构建的行业适用性范围有限。[结果/结论]构建维度分别是:味道、环境、服务、菜品、价格,画像维度得分:5.0、2.13、2.1、1.31、1.0。实验结果证实,该方法能够有效构建店面画像,且构建的各维度得分与语料库得分具有一致性,实验表明该构建方法能取得良好效果。

[Objective/Significance]We construct the store portrait evaluation dimensions and store portrait score from the Meituan catering user review data to help stores improve the level of digital and intelligent management,realize in-depth user value mining and maximize store revenue.[Methods/Process]Firstly,keywords are extracted for BERT word vector representation,and the evaluation dimension of storefront portrait is constructed through Gaussian mixture dimension clustering,dimension recognition and dimension word mining,and then the domain sentiment dictionary and syntactic analysis are constructed to obtain dimension word scores,and finally the constructed store portraits are scored.[Limitations]The disadvantage is that this experiment only focuses on the reviews of gourmet buffets in Wuhan,and the scope of industry applicability of its construction is limited.[Results/Conclusions]The construction dimensions include:taste,environment,service,dishes,and price,and the portrait dimension scores are 5.0,2.13,2.1,1.31 and 1.0.Experimental results confirm that the proposed method can effectively construct the storefront portrait,and the scores of each dimension of the construction are consistent with the scores of the corpus,and the experiment shows that the construction method can achieve good results.