Zhenhua Meng · ACM International Conference on Information and Knowledge Management (CIKM) 2026 · 2026
DOI: 10.1145/3799682.3840717
Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).
Multi-criteria ratings provide user feedback on multiple aspects of an item, offering richer preference signals than a single overall rating. Early work mainly focused on predicting criterion or overall ratings, whereas practical scenarios require ranking items according to their overall utility inferred from sparse and heterogeneous criterion-level evidence, which constitutes the core problem of multi-criteria recommendation (MCR). This ranking-oriented formulation of MCR gives rise to two key challenges: learning reliable criterion dependencies under sparse interactions, and bridging the mismatch between semantic preference signals and collaboratively learned representations. To address these issues, we propose S2R-LMG, a scoring-to-ranking framework that jointly learns from semantic and collaborative views. Specifically, we construct a user–item–criterion tripartite graph to model criterion dependencies in the collaborative view, while using large language models (LLMs) to transform sparse multi-criteria feedback into criterion-aware contextual preference priors for users and items. These priors are then projected into the ranking space and coupled with collaborative representations via alignment and retrieval-based distillation, so that semantic guidance directly contributes to preference ranking. Extensive experiments on three benchmark datasets show that S2R-LMG consistently outperforms state-of-the-art models on top-N ranking, achieving an average improvement of 4.84% across all datasets and evaluation metrics over the strongest baseline. The implementation code is available at https://github.com/nunu1995/S2R-LMG.
ACM International Conference on Information and Knowledge Management (CIKM) 2026 · 0 citations
ACM International Conference on Information and Knowledge Management (CIKM) 2026 · 0 citations
No comments yet — start the discussion below.