With the rapid digitization of the art industry, the demand for intelligent artwork recommendation systems that understand stylistic compatibility has grown substantially. However, existing methods often focus on individual artwork similarity while overlooking higher-order correlations among multiple artworks and the distributed nature of modern recommendation environments. This paper proposes a novel framework, Style-Aware Artwork Recommendation with Cloud-Edge Collaborative Knowledge Graphs (SAR-CE-KG), which recommends a set of mutually compatible artworks that collectively fulfill a user’s stylistic preferences. Specifically, we construct an artwork knowledge graph, where nodes represent artworks annotated with style descriptors and edges capture historical co-occurrence frequencies, reflecting their stylistic compatibility. Given a user query composed of target style keywords, the system identifies a connected subgraph—representing a style-compatible artwork set—whose combined style coverage maximizes both completeness and internal harmony. Cloud servers perform global knowledge graph aggregation and compatibility modeling, while edge devices enable real-time, privacy-preserving personalization. Experimental evaluations on real-world artwork collections demonstrate that SAR-CE-KG achieves superior performance in style coverage, recommendation coherence, and response latency, highlighting the effectiveness of cloud–edge synergy for large-scale intelligent art recommendation.
Li et al. (Tue,) studied this question.