Content processing method and apparatus, computer device, and storage medium

A computer device, comprising a processor and a memory, the memory storing computer-readable instructions, the processor being configured to execute the computer-readable instructions to cause the computer device to:

• obtain a description text describing a totality of a to-be-processed content comprising a plurality of video frames;
• extract an image from the to-be-processed content, wherein the image is a key image extracted among the plurality of video frames of the to-be-processed content, and wherein the extracting comprises:
• extracting at least two video key frames from the to-be-processed content;
• determining a relevance between each of the at least two video key frames and the description text, wherein the relevance is between a key action in a movement or a change of character or an object in the key image and the description text;
• and
• determining the key image in the at least two video key frames according to the relevance;
• perform feature extraction on the description text based on text semantics to obtain a text eigenvector;
• perform feature extraction on the image based on image semantics to obtain an image eigenvector;
• combine the text eigenvector with the image eigenvector to obtain an image-text multi-modal vector;
• and
• generate an estimated click-through rate of the to-be-processed content according to the image-text multi-modal vector.

A non-transitory computer-readable storage medium, storing computer-readable instructions, the computer-readable instructions being executed by one or more processors to cause a computer device to performed steps, comprising:

• obtaining a description text describing a totality of a to-be-processed content comprising a plurality of video frames;
• extracting an image from the to-be-processed content, wherein the image is a key image extracted among the plurality of video frames of the to-be-processed content, and wherein the extracting comprises:
• extracting at least two video key frames from the to-be-processed content;
• determining a relevance between each of the at least two video key frames and the description text, wherein the relevance is between a key action in a movement or a change of character or an object in the key image and the description text;
• and
• determining the key image in the at least two video key frames according to the relevance;
• performing feature extraction on the description text based on text semantics to obtain a text eigenvector;
• performing feature extraction on the image based on image semantics to obtain an image eigenvector;
• combining the text eigenvector with the image eigenvector to obtain an image-text multi-modal vector;
• and
• generating an estimated click-through rate of the to-be-processed content according to the image-text multi-modal vector.