Television content representation method and device, equipment, storage medium and product

The invention discloses a television content representation method and device, equipment, a storage medium and a product, and the method comprises the steps: carrying out the coding of each piece of modal information of television content, and obtaining a graph representation corresponding to each piece of modal information; performing feature interaction of nodes in the graph based on the graph representation to obtain a feature set of the nodes in the graph; performing inter-graph feature interaction based on the graph representation and the intra-graph node feature set to obtain inter-graph feature interaction information; and fusing the feature interaction information among the graphs to obtain television content representation. And the accuracy of television content representation is improved. 本申请公开了电视内容的表示方法、装置、设备、存储介质及产品，该方法包括：对电视内容的各模态信息分别进行编码，得到各模态信息对应的图表示；基于所述图表示进行图内节点的特征交互，得到图内节点特征集合；基于所述图表示和所述图内节点特征集合进行图间特征交互，得到图间特征交互信息；对各个图间特征交互信息进行融合，得到电视内容表示。提高电视内容表示的准确性。