计算机科学
推荐系统
二部图
情报检索
熵(时间箭头)
骨料(复合)
图形
协同过滤
万维网
重新使用
情绪分析
微博
数据挖掘
最大熵原理
基本事实
截断(统计)
语义相似性
机器学习
主题模型
数据科学
冷启动(汽车)
用户满意度
用户生成的内容
在线广告
人工智能
社会化媒体
电子商务
作者
Guofang Nan,Ruixuan Wang,Jiwei Luo,Zhiyun Ji
标识
DOI:10.1080/10864415.2026.2731859
摘要
Online travel platforms aggregate a vast number of online reviews, reflecting users’ diverse attitudes and opinions about products. However, existing recommendation methods fail to accurately capture user preferences from these reviews, making it difficult to provide personalized recommendations. To overcome these limitations, we propose a novel information entropy-based hotel recommendation method based on online reviews to provide users with accurate recommendations. Specifically, aspect–opinion pairs and their sentiment polarities are first extracted from reviews. Then, sentiment entropy and semantic entropy are constructed to capture user preferences. Next, these features are incorporated into a user–hotel bipartite graph, where node2vec is applied to learn node representations. Last, the graph features, user features and hotel features are inputted into a BiGRU model for recommendation. Numerical experiments show that our method produces more accurate recommendations than several well-known baselines. Furthermore, we propose a semantic entropy-based review truncation method that significantly reduces computational costs while preserving recommendation quality.
科研通智能强力驱动
Strongly Powered by AbleSci AI