JevRec improves item recommendation by encoding popular items with shorter codes
JevRec: Linguistic Encoding with Jev-like Decision Modeling for LLM-based Generative Recommendation
Computation and Language
Summary
Recommender systems suggest items to users by predicting what they might want next. This work improves how items are represented in the recommendation process by giving frequent items shorter, simpler codes and less common items longer, richer codes, similar to how common words in language are shorter. The authors also separate the prediction of code length from code content to make the system more efficient. Their approach, called JevRec, showed better recommendations and faster training and prediction compared to previous methods.
What this means in practice
- •For e-commerce platform engineers: Enhance product recommendation systems by encoding popular products with shorter codes to increase speed and improve accuracy.
- •For streaming service developers: Improve personalized content suggestions by using frequency-aware item encoding that better captures user preferences.
Authors
Qiyong Zhong, Jiajie Su, Ming Yang, Li Zhang
Abstract
Generative recommendation (GR) represents items as semantic identifiers (SIDs) and autoregressively generates the next item's SID from interaction history. However, most methods assign the same SID length to all items despite highly unequal interaction frequencies. This uniform design repeatedly incurs the cost of fine-grained codes for popular items, while shortening codes to reduce this cost also compresses the semantic representations of long-tail items. Inspired by the compact forms of frequent expressions in natural language, we propose JevRec to encode items as variable-length words and generate the next item from interaction context. Linguistic encoding retains frequency-dependent semantic prefixes from shared codebooks, producing shorter SIDs for popular items and richer descriptions for long-tail items. Yet predicting a sequence-level length and generating token-level content impose different optimization requirements that may conflict under a shared output head. Our Jev-like decision modeling component therefore predicts a length distribution with a dedicated head and generates content conditioned on length through a separate head while sharing the interaction-history representation. Experiments on three public datasets show that JevRec outperforms all compared baselines across the evaluated recommendation metrics. Further analyses show lower training and inference time than TIGER and reduced gradient conflict between length and content prediction.