Video Recommendation Using Content Characteristic Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video recommendation methods primarily rely on video names, failing to effectively suggest videos similar in content to the user's desired video.
Innovation Solution
A method and apparatus that recommend videos by finding similarities in content characteristics, using pre-trained neural networks like bidirectional Long Short-Term Memory (LSTM) to extract and compare content characteristic vectors, and sending play information of similar videos to the user's terminal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If video recommendation is based on video names, then the recommendation method is simple to implement, but the recommendation accuracy and content similarity are poor
Solution Approach 1:
The patent transforms the recommendation basis from video names (text parameter) to video content characteristics (visual and semantic parameters). By extracting content characteristics including object types, actions, and semantic information from video frames, the system changes the fundamental parameter used for similarity calculation, thereby improving recommendation accuracy while maintaining computational feasibility through efficient feature extraction methods.
Solution Approach 2:
The patent introduces content characteristic vectors as an intermediary between video content and recommendation algorithms. Instead of directly comparing video names or raw video data, the system extracts meaningful content characteristics (objects, actions, semantics) and represents them as vectors, which then serve as the basis for similarity calculation. This intermediary representation bridges the gap between complex video content and simple recommendation metrics.
2Measurement precision
If video recommendation is based on content characteristics using neural networks, then the recommendation accuracy improves, but the computational complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-training neural network models to extract content characteristics from video frames before the actual recommendation process. The system pre-extracts and stores content characteristic vectors (including object types, actions, and semantic information) for videos, so that during recommendation, only similarity calculation is needed rather than full video analysis. This pre-processing step reduces real-time computational complexity while maintaining high recommendation accuracy.
Solution Approach 2:
The patent segments video content analysis into distinct components: object detection, action recognition, and semantic extraction. Each component is handled by specialized neural network modules that process different aspects of video content independently. This segmentation allows the system to compute content characteristics in a modular fashion, reducing overall computational complexity compared to analyzing entire video sequences as a single unit.
Data Source
AI summary
Embodiments of the present disclosure disclose a method and apparatus for recommending a video. A specific implementation of the method includes: finding a recommended video corresponding to a target video from all candidate videos based on similarities of content characteristics of the videos, the target video being a video to be played on a terminal of a user; and sending play information of the recommended video corresponding to the target video to the terminal of the user. The method finds a video similar on video content to the target video that the user desires to view based on the content characteristic of the video, and recommends the video similar on video content to the target video that the user desires to view to the user.