Content recommendation methods, devices, electronic devices, and computer-readable storage media

Abstract

This invention discloses a content recommendation method, apparatus, electronic device, and computer-readable storage medium. After acquiring a set of recommended object samples, a set of content samples, and interaction information between recommended object samples and content samples, this invention extracts features from the recommended object samples and content samples to obtain recommendation features. Based on the interaction information, it constructs an interaction graph network using the recommended object samples and content samples as nodes, propagates the recommendation features within the interaction graph network, and calculates the content distance between content samples in the content sample set. Based on the propagated recommendation features and content distances, it selects recommended content samples corresponding to each recommended object from the content sample set. The interaction graph network is then updated based on the recommended content samples to obtain a trained interaction graph network, which is then used to recommend content to the objects to be recommended. This scheme can improve the accuracy of content recommendation.

Images