可解释推荐系统中的解释风格研究综述OA

This paper systematically analyzes explanation style mechanisms in explainable recommendation systems from four dimensions:user,item,content generation,and external knowledge.For each explanation style,this paper first elaborates on its basic concepts and core ideas,then analyzes its implementation approaches,key strategies,and technical methods from specific types.Finally,a comprehensive comparison of various explanation styles and methods is conducted from aspects of core mechanisms,technical support,main advantages,and potential limitations.The research shows that different explanation styles have varying impacts on user understanding and acceptance,requiring the selection of appro-priate styles based on user preferences and system objectives.Moreover,certain explanation styles can reveal the opera-tional mechanisms and decision logic of recommendation algorithms to some extent.The effectiveness of different expla-nation styles in achieving specific explanation objectives has not been evaluated,nor has the enhancement effect of expla-nation styles on user trust been examined.Additionally,influencing factors such as the number of explanations and presen-tation order(such as the ranking of users'preferred explanation styles)have not been considered.Although existing expla-nation styles have made significant progress in meeting specific application needs in different domains,they still face num-erous challenges in aspects such as constituent feature impacts,scenario differences and user cognitive investment willing-ness,structured argumentation,and offline evaluation.