Does noise in the knowledge graph really harm recommendations?

Meng Yan

doi:10.1016/j.patcog.2025.112657

Does noise in the knowledge graph really harm recommendations?

2025-11-08

0

PRE

AI

DOI：10.1016/j.patcog.2025.112657

Original

Original request for help

Save

Abstract

En 中文

• We propose to mine user preferences from a noisy knowledge graph, synthesize topological and semantic information to evaluate the confidence of triples. • We point out that noise is not entirely useless and give definitions of useful noise and useless noise. Moreover, we design a kernel function to model positive and negative preference signals uniformly. • We conduct extensive experiments on the available real-world datasets, and experimental results show that the proposed method can significantly improve performance and outperform state-of-the-art methods.

Does noise in the knowledge graph really harm recommendations?

Abstract

Journal

Organization