《计算机应用》唯一官方网站 ›› 2023, Vol. 43 ›› Issue (S1): 88-94.DOI: 10.11772/j.issn.1001-9081.2022071087

• 人工智能 • 上一篇    下一篇

基于文本挖掘的物流服务水平评价方法

陈清化, 薛书琦(), 龚壮壮, 曹润康   

  1. 西安邮电大学 现代邮政学院,西安710061
  • 收稿日期:2022-07-27 修回日期:2022-11-03 接受日期:2022-11-05 发布日期:2023-01-15 出版日期:2023-06-30
  • 通讯作者: 薛书琦
  • 作者简介:陈清化(1997—),女,山西临汾人,硕士研究生,主要研究方向:物流服务水平评价;
    薛书琦(1990—),男,甘肃平凉人,副教授,博士,主要研究方向:交通行为分析、车辆路径规划;shuqixue@xupt.edu.cn
    龚壮壮(1997—),男,河南驻马店人,硕士研究生,主要研究方向:车辆路径规划;
    曹润康(1998—),男,山西长治人,硕士研究生,主要研究方向:大数据商务分析。

Evaluation method of logistics service level based on text mining

Qinghua CHEN, Shuqi XUE(), Zhuangzhuang GONG, Runkang CAO   

  1. School of Modern Posts,Xi’an University of Posts and Telecommunications,Xi’an Shaanxi 710061,China
  • Received:2022-07-27 Revised:2022-11-03 Accepted:2022-11-05 Online:2023-01-15 Published:2023-06-30
  • Contact: Shuqi XUE

摘要:

物流服务水平评价是物流企业及电商平台提高客户满意率、增加顾客忠诚度的重要基础。传统方法一般是基于评价指标体系结合权重设计对物流服务水平进行评价,这些方法往往存在指标不全面、评价过程主观性强的问题。为解决这些问题,提出基于文本情感计算模型的物流服务水平评价方法。在现有开源词典基础上,加入网络情感词典,通过词频-逆文档频率(TF-IDF)关键词提取算法及词频排序建立适用于物流服务评价的词典。语义规则部分,考虑句间关系和句型关系,优化情感单元及语义计算规则,从词语级、分句级到复句级的角度依次对物流服务关键句评论进行情感得分计算,根据得分作出评价。以京东和淘宝平台内电子产品、服装及生鲜3类产品的评论文本为对象进行实验验证。实验结果表明:1)所提方法在精确率、召回率及F值指标上均明显优于支持向量机(SVM)、卷积神经网络(CNN)分类的结果,分别平均提高约17%和33%;且比单独使用物流领域情感词典或优化后计算规则的结果分别平均提高约17%和18%;2)综合3类产品的评论,京东的平均物流服务满意率高于淘宝,两个平台的3类产品中生鲜类商品的平均物流服务满意率最高,服装类的平均满意率最低。

关键词: 物流评论, 情感词典, 语义规则, 情感分析, 物流服务水平

Abstract:

Evaluation of logistics service level is an important foundation for logistics enterprises and e-commerce platforms to improve the satisfaction and loyalty of their customers. Traditional methods are usually based on evaluation index system design combined with weight determination, which are criticized for incomplete indexes and subjectivity in evaluation. To address these problems, a logistics service level evaluation method based on a text motion computing model was proposed. With existing open source dictionary, a network emotion dictionary was added, and a dictionary suitable for logistics service evaluation was established through the keyword extraction algorithm of Term Frequency-Inverse Document Frequency (TF-IDF) and word frequency sorting. In semantic rule section, emotional units and semantic calculation rules were improved by considering the sentence type relationship and relationships between sentences. The sentiment score of the key sentences related to logistics service was evaluated from the perspective of word level, clause level, and complex sentence level. The validation was conducted on the comments of electronic products, clothing and fresh food from two platforms of JD and Taobao. Experimental results show that: 1) The classification results of the proposed method were significantly better than those of Support Vector Machine (SVM) as well as those of Convolutional Neural Network (CNN) in terms of the average of precison, recall and F value indicators, with an average improvement of about 17% and 33%, respectively; and compared with the results obtained by using only the sentiment dictionary of logistics or the optimized calculation rules, the average of precison, recall and F value increased by about 17% and 18%, respectively. 2) Based on comments of the three categories of products, the average logistics service satisfaction rate of JD was higher than that of Taobao. Among the three categories of products from JD and Taobao, the average logistics service satisfaction rate of fresh food products was the highest, and the average satisfaction rate for clothing products was the lowest.

Key words: logistics comment, emotional dictionary, semantic rule, sentiment analysis, logistics service level

中图分类号: