From Side Information and Knowledge Graphs to Large Language Models: Two Decades of Knowledge Integration in Recommender Systems

Abstract

Over the past two decades, Recommender Systems (RSs) have repeatedly changed how they integrate external knowledge, moving from constraints and side information to Knowledge Graphs (KG) and, more recently, Large Language Models (LLMs). This Past/Present/Future retrospective interprets that evolution as a sequence of representational translations rather than simple replacements. Across representative RecSys contributions, we identify recurring patterns: external knowledge is repeatedly used to address sparsity and cold start, human-interpretable structure remains central to explanation and control, and hybrid architectures re-emerge whenever scalability conflicts with traceability. We argue that LLMs should be understood not as substitutes for structured knowledge, but as a new interaction layer that requires explicit grounding, and we conclude with a cautious agenda for hybrid, inspectable, and accountable RSs.