An Aligning and Training Framework for Multimodal Recommendations

arXiv:2403.12384v1 Announce Type: cross
Abstract: With the development of multimedia applications, multimodal recommendations are playing an essential role, as they can leverage rich contexts beyond user interactions. Existing methods mainly regard multimodal information as an auxiliary, using them to help learn ID features; however, there exist semantic gaps among multimodal content features and ID features, for which directly using multimodal information as an auxiliary would lead to misalignment in representations of users and items. In this paper, we first systematically …

abstract applications arxiv beyond cs.ir cs.lg development features framework however information interactions learn multimedia multimodal multimodal content playing recommendations regard role semantic them training type