Video Recommendation System Segmentation for Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video recommending methods do not accurately differentiate between long and short videos, leading to suboptimal recommendation accuracy due to uniform processing of both formats.
Innovation Solution
A method that separates long and short videos based on attribute thresholds and user watching records, recommending videos with high attribute values and similarity matches to enhance personalization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If long videos and short videos are processed uniformly in conventional recommendation methods, then the system complexity is reduced, but the recommendation accuracy deteriorates
Solution Approach 1:
The patent divides the video recommendation system into two separate processing paths: one for long videos and one for short videos. Each path applies different recommendation strategies tailored to the specific characteristics of that video type, thereby improving recommendation accuracy without excessively increasing system complexity through modular design.
Solution Approach 2:
The patent applies different recommendation algorithms and criteria to different video types. For long videos, it uses one set of features and sorting criteria, while for short videos, it uses another set optimized for that format. This localized optimization improves overall recommendation quality.
2Measurement precision
If separate processing paths are created for long and short videos, then recommendation accuracy is improved, but the device complexity increases
Solution Approach 1:
The system is segmented into distinct processing modules for long videos and short videos, each with specialized recommendation logic. This modular segmentation improves accuracy while keeping complexity manageable through clear separation of concerns.
Solution Approach 2:
The patent employs a unified recommendation framework that can handle both long and short videos, with configurable parameters and algorithms that adapt to different video types. This multi-functional approach improves accuracy without requiring completely separate systems.
3Measurement precision
If attribute thresholds are applied to filter long videos, then the quality of recommended long videos is improved, but the quantity of available videos decreases
Solution Approach 1:
The patent dynamically adjusts attribute thresholds based on user preferences, watching history, and video availability. By changing the parameter values of thresholds adaptively, the system maintains high video quality while ensuring sufficient video quantity for recommendations.
Solution Approach 2:
The system applies attribute thresholds selectively to ensure minimum quality standards are met, rather than applying overly strict filters that would eliminate too many videos. This partial application of filtering maintains quality while preserving adequate video quantity.
Data Source
AI summary
A video recommending method, including: obtaining videos, the video including long videos and short videos; obtaining a subset of the long videos, of which video attribute values are greater than corresponding attribute thresholds; obtaining a watching record of a user, and obtaining similarities between the short videos and videos in the watching record, to extract a preset quantity of short videos having highest similarities; and recommending the subset of the long videos, of which video attribute values are greater than corresponding attribute thresholds, to the user, and/or recommending the preset quantity of short videos having highest similarities to the user.