Personalized Review Summarization via Online Preference Learning
A new framework called PREFER generates personalized product review summaries by learning user preferences online. Unlike static summarization systems, PREFER adapts to individual users' evolving interests through iterative feedback from generated summaries. A case study on the Amazon Reviews'23 dataset demonstrates improved alignment with target user interests while maintaining summary quality. The approach addresses the challenge of unknown latent preferences in e-commerce, where different users care about different product characteristics and these preferences change with interactions.
Key facts
- PREFER is an online learning framework for personalized review summarization.
- It iteratively refines user preferences via feedback from generated summaries.
- A case study uses the Amazon Reviews'23 dataset.
- Controlled simulations show improved alignment with target user interests.
- The system addresses unknown latent preferences.
- Current systems produce generic, static summaries.
- User preferences may evolve with interactions.
- The framework generates personalized summaries for each user.
Entities
Institutions
- Amazon