Dynamic Tag List for Mobile Content Recommendation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing content recommendation methods for mobile terminals often push irrelevant content to users, failing to accurately identify and prioritize content of interest, leading to a poor user experience due to missed relevant content and continued exposure to uninteresting material.
Innovation Solution
A method where a server generates a tag list based on user browsing history and previously selected tags, allowing users to select their interests, and pushes content corresponding to these tags, ensuring more accurate and relevant content delivery.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If existing recommendation methods push content based on initial user preferences, then content delivery automation is improved, but content recommendation accuracy deteriorates because users lose interest over time
Solution Approach 1:
The patent implements dynamic tag lists that are automatically updated based on user feedback. The system transitions from static initial preferences to dynamic evolving preferences by continuously adding selected tags and removing unselected tags from the tag list, allowing the recommendation system to adapt to changing user interests over time
Solution Approach 2:
The patent incorporates user feedback mechanisms where user selections and non-selections are processed to update the tag list. The server receives feedback about which tags users select and which they don't, then automatically adjusts the tag list accordingly, creating a closed-loop system that improves recommendation accuracy through continuous learning from user behavior
2Productivity
If the system pushes content based on initial tags, then push automation is improved, but user experience deteriorates due to continued exposure to uninteresting content
Solution Approach 1:
The tag list evolves dynamically based on user interactions. As users select or reject content tags, the system automatically updates the tag list to reflect current user interests, ensuring that pushed content remains relevant and improving user experience while maintaining automation
Solution Approach 2:
The system discards tags that users consistently reject and recovers or adds tags that users select. This process of eliminating unwanted content categories and recovering preferred ones ensures that the pushed content aligns with user interests, improving ease of operation and user experience
3Device complexity
If the system uses simple browsing history analysis, then device complexity is reduced, but recommendation completeness deteriorates due to missed relevant content
Solution Approach 1:
The patent introduces tags as intermediary elements that bridge user browsing behavior and content recommendations. Instead of directly analyzing complex browsing patterns, the system uses tags as a simplified intermediary representation that captures user interests while maintaining recommendation completeness with lower complexity
4Measurement precision
If the system continuously updates recommendations, then recommendation accuracy is improved, but information loss increases due to filtering out potentially relevant content
Solution Approach 1:
The system performs partial updates to the tag list by only adding or removing specific tags based on user feedback rather than completely replacing the tag list. This partial action approach maintains recommendation accuracy by focusing on changing user preferences while preserving the broader content coverage through retained tags
Data Source
Figure 1~3
Figure 4
Figure 5
AI summary
The present application discloses a content recommendation method and device, a mobile terminal, and a server, relating to the technical field of the Internet. The method comprises: obtaining a first label list, the first label list being generated according to a label corresponding to content browsed by a user within a target time period; obtaining a second label list, the second label list being generated according to both the first label list and a saved recommendation label list; obtaining a label selected by the user from the second label list to generate a new recommendation label list; and pushing to the mobile terminal content corresponding to a label in the new recommendation label list. With the solution, content that a user is interested in can be pushed and content that the user is not interested in can be removed, thereby achieving a better pushing effect.