Explainable Paper Recommendations Based on Heterogeneous Graph Representation Learning and the Attention Mechanism
Paper-recommender systems aim to recommend the most relevant academic papers for researchers from the vast academic resources.Existing paper recommendation methods are mainly based on the textual information of papers or citation relationships,failing to make full use of the rich semantic information in heterogeneous academic graphs,thus leading to less accurate recommendation results.In addition,existing methods focus more on recommendation accuracy and neglect interpretability,which decreases the reliability and user satisfaction of the paper recommender systems.To solve these issues,this study proposes an Explainable Paper Recommendation(EPRec)method based on heterogeneous graph representation learning and the attention mechanism.First,an attention mechanism-enhanced heterogeneous graph representation learning-based paper recommendation module is proposed to incorporate multi-source side information from heterogeneous academic graphs.Then,the interpretable text generation method is introduced into the paper recom-mendation scenario.A feature-based text generation module is proposed to generate textual interpretation to explicitly pro-vide recommendation reasons to researchers.Finally,we construct an academic dataset that contains multi-source informa-tion,including paper metadata,feature words,and citation contexts.The experimental results show that EPRec performs better than the comparative methods in terms of precision and recall.Moreover,EPRec provides high-quality,interpretable text explanations for the paper recommendation results.
paper recommendationheterogeneous graph representation learningattention mechanisminterpretable text generationmeta-path
万方数据
维普
