PREFER：基于在线偏好学习的个性化评论摘要
PREFER: Personalized Review Summarization with Online Preference Learning
AI 导读
PREFER 提出一种在线学习框架，通过在潜在 aspect 空间中表示用户兴趣并选择符合偏好的评论证据，为用户生成个性化商品评论摘要，并随反馈迭代优化偏好理解。该框架在 Amazon Reviews'23 上用受控反馈模拟及 held-out-history、独立嵌入、外部 LLM 评估器做验证，偏好对齐的证据选择表现提升。人类研究显示，场景化摘要在 69.9% 的配对比较中优于通用汇总摘要。
正文 · 原文
Abstract:Product reviews significantly influence purchasing decisions on e-commerce platforms. However, the sheer volume of reviews can overwhelm users, obscuring information most relevant to their individual needs. Existing e-commerce summarization systems typically produce generic, static summaries that fail to account for the fact that (i) different users prioritize different product characteristics, and (ii) these preferences may evolve with interactions. To address these challenges, we propose an online learning framework that generates personalized summaries for each user by representing user interests in a latent aspect space and selecting preference-aligned review evidence. Our system then iteratively refines its understanding of user preferences by incorporating feedback on the generated summaries directly over time. We evaluate our framework on Amazon Reviews'23 using controlled feedback simulations and robustness checks with held-out-history, independent-embedding, and external LLM-based evaluators. These experiments show improvements in preference-aligned evidence selection under the evaluated feedback settings. A separate human-based study finds that scenario-conditioned summaries are preferred over generic (uniformly-aggregated) alternatives in 69.9% of paired comparisons.
| Subjects: | Artificial Intelligence (cs.AI); Computer Science and Game Theory (cs.GT); Machine Learning (cs.LG); Systems and Control (eess.SY); Optimization and Control (math.OC) |
| Cite as: | arXiv:2605.05911 [cs.AI] |
| (or arXiv:2605.05911v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2605.05911 arXiv-issued DOI via DataCite |
Submission history
From: Millend Roy [view email]
[v1]
Thu, 7 May 2026 09:20:47 UTC (2,164 KB)
[v2]
Mon, 26 Oct 2026 03:47:06 UTC (615 KB)
来源：arXiv cs.AI · arxiv.org