Media content processing method and apparatus, storage medium, and electronic device

An electronic device comprising:

• at least one memory storing one or more instructions;
• and
• at least one processor communicatively connected to the at least one memory and configured to execute the one or more instructions to:
• obtain a first heterogeneous graph, including constructing a first meta-path, the first meta-path including a first start information node representing historical query information and at least one first other information node representing media information, the media information including one media content and associated information about the media content, and the media content being in a query result corresponding to the historical query information and being operated by a user;
• obtain a second heterogeneous graph, including constructing a second meta-path, the second meta-path including a second start information node representing the media content and at least one second other information node representing the query information or the associated information;
• train a feature extraction network, based on the first heterogeneous graph and the second heterogeneous graph, to obtain a graph feature extraction model, comprising:
• constructing a plurality of positive sample pairs and negative sample pairs based on a plurality of pieces of historical query information and media content corresponding to each of the plurality of pieces of historical query information;
• for each sample pair in the plurality of positive sample pairs and the plurality of negative sample pairs, performing:
• extracting, based on the first heterogeneous graph and the second heterogeneous graph, respectively, a first sub-graph corresponding to historical query information in the sample pair and a second sub-graph corresponding to media content corresponding to the historical query information;
• and
• separately inputting the first sub-graph and the second sub-graph into the feature extraction network to perform graph feature extraction, to obtain first sample representation information corresponding to the historical query information and second sample representation information corresponding to the media content corresponding to the historical query information;
• and
• adjusting a parameter of the feature extraction network based on a loss, to obtain the graph feature extraction model, the loss being configured to guide the feature extraction network to enhance a correlation between first sample representation information and second sample representation information in each of the positive sample pairs and reduce a correlation between first sample representation information and second sample representation information in each of the negative sample pairs;
• and
• input the media content into the graph feature extraction model to obtain representation information corresponding to the media content.

A non-transitory computer-readable storage medium storing at least one instruction or at least one program that, when loaded and executed by at least one processor, causes the at least one processor to:

• obtain a first heterogeneous graph, including constructing a first meta-path, the first meta-path including a first start information node representing historical query information and at least one first other information node representing media information, the media information including one media content and associated information about the media content, and the media content being in a query result corresponding to the historical query information and being operated by a user;
• obtain a second heterogeneous graph, including constructing a second meta-path, the second meta-path including a second start information node representing the media content and at least one second other information node representing the query information or the associated information;
• and
• determine and store, train a feature extraction network based on the first heterogeneous graph and the second heterogeneous graph, to obtain a graph feature extraction model, comprising:
• constructing a plurality of positive sample pairs and negative sample pairs based on a plurality of pieces of historical query information and media content corresponding to each of the plurality of pieces of historical query information;
• for each sample pair in the plurality of positive sample pairs and the plurality of negative sample pairs, performing:
• extracting, based on the first heterogeneous graph and the second heterogeneous graph, respectively, a first sub-graph corresponding to historical query information in the sample pair and a second sub-graph corresponding to media content corresponding to the historical query information;
• and
• separately inputting the first sub-graph and the second sub-graph into the feature extraction network to perform graph feature extraction, to obtain first sample representation information corresponding to the historical query information and second sample representation information corresponding to the media content corresponding to the historical query information;
• and
• adjusting a parameter of the feature extraction network based on a loss, to obtain the graph feature extraction model, the loss being configured to guide the feature extraction network to enhance a correlation between first sample representation information and second sample representation information in each of the positive sample pairs and reduce a correlation between first sample representation information and second sample representation information in each of the negative sample pairs;
• and
• input the media content into the graph feature extraction model to obtain representation information corresponding to the media content.