A Review-based Transformer Model for Personalized Product Search
share
In product search, customers make purchase decisions based on not only the product relevance but also their personal preferences. Despite its great potential, recent analysis on commercial logs shows that personalization does not always improve the quality of product search. Most existing personalized product retrieval models, however, cannot control whether to apply personalization or not under different contexts. From the perspective of model design, the typical paradigm of existing models is to represent and match user intents and items in the semantic space, where finer-grained matching is totally discarded and the ranking of an item cannot be explained further than just user/item level similarity. In addition, while some models in existing studies have created dynamic user representations based on search context, their representations for items are static across all search sessions. This makes every piece of information about the item always have equal importance in representing the item during matching with various user intents. Given the above limitations,we propose a review-based transformer model (RTM) for personalized product search, which encodes the sequence of query, user reviews, and item reviews with a transformer architecture. RTM can perform adaptive personalization with different accumulative attention weights on user reviews in various search sessions. Moreover, review-level matching is conducted between the user and item, where each reviewhas a dynamic effect according to the context in the sequence. This makes it possible to identify useful reviews to explain the scoring. Experimental results show that RTM achieves significantly better performance than state-of-the-art personalized product search baselines.
READ FULL TEXT
