Scholarlink

The review text in hotel search systems is beneficial for the general public searching for hotels. However, the benefits of posting reviews for reviewers themselves are limited, and the number of users posting reviews is not large. Therefore, we thought that we could increase the number of users posting reviews by developing a system that provides useful information for reviewers. Based on the six evaluation items commonly found on conventional hotel reservation sites, we created 19 new evaluation items aimed at eliminating user dissatisfaction. For each of these evaluation items, we constructed a model that classifies reviews into positive, negative, and unmentionable reviews using word embedding technology. This model allows us to analyze which evaluation items the reviewers express dissatisfaction with and which they express satisfaction with. It then predicts a score for each evaluation item that indicates the likelihood of eliminating the reviewer’s dissatisfaction, and recommends hotels that exceed the predicted score. This paper proposes several methods for calculating reviewer dissatisfaction resolution scores and examines dissatisfaction resolution methods through evaluation experiments. We also examine ranking methods for presenting recommended hotels to users.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An Examination of Methods for Resolving Reviewer Dissatisfaction in Hotel Recommendation Systems

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Yujiro Hayashi,
  • Jianwei Zhang,
  • Yukiko Kawai,
  • Shinsuke Nakajima

摘要

The review text in hotel search systems is beneficial for the general public searching for hotels. However, the benefits of posting reviews for reviewers themselves are limited, and the number of users posting reviews is not large. Therefore, we thought that we could increase the number of users posting reviews by developing a system that provides useful information for reviewers. Based on the six evaluation items commonly found on conventional hotel reservation sites, we created 19 new evaluation items aimed at eliminating user dissatisfaction. For each of these evaluation items, we constructed a model that classifies reviews into positive, negative, and unmentionable reviews using word embedding technology. This model allows us to analyze which evaluation items the reviewers express dissatisfaction with and which they express satisfaction with. It then predicts a score for each evaluation item that indicates the likelihood of eliminating the reviewer’s dissatisfaction, and recommends hotels that exceed the predicted score. This paper proposes several methods for calculating reviewer dissatisfaction resolution scores and examines dissatisfaction resolution methods through evaluation experiments. We also examine ranking methods for presenting recommended hotels to users.