Sticker Recommendation System Using Emotion Feature Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current sticker recommendation systems on internet platforms often fail to suggest suitable sticker sets to users, as they primarily rely on popularity, leading to unsuitable recommendations and wastage of resources, as different users have varying preferences for different stickers within a set.
Innovation Solution
A method and apparatus that recommend sticker sets by analyzing historical sticker usage, calculating emotion features, and determining recommendation indexes based on similarities and user preferences, using techniques such as weighted summation and clustering to suggest suitable sticker sets.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If sticker sets are recommended based on usage popularity, then the recommendation system is simple to implement, but the recommendation relevance to individual users deteriorates
Solution Approach 1:
The patent transforms the recommendation basis from a single parameter (usage popularity) to multiple parameters including emotion features, user preferences, and sticker set characteristics. This allows the system to maintain simplicity while improving recommendation relevance by changing what parameters are considered.
Solution Approach 2:
The system performs preliminary analysis of user historical sticker usage patterns and emotion features before making recommendations. By pre-processing user data and sticker set attributes, the system prepares personalized recommendation criteria in advance, improving relevance without adding complex real-time processing.
2Speed
If sticker sets are recommended based on usage popularity, then the recommendation process is fast, but resource wastage increases due to unsuitable recommendations
Solution Approach 1:
The system enables users to self-evaluate and provide feedback on recommended sticker sets through explicit feedback mechanisms. This user-driven evaluation helps the system learn from actual user preferences, reducing resource wastage by focusing on sticker sets that users genuinely find suitable rather than just popular ones.
Solution Approach 2:
The patent implements feedback loops where user interactions with recommended sticker sets (usage, ratings, preferences) are continuously collected and used to refine future recommendations. This feedback mechanism improves resource utilization by directing recommendations toward sticker sets that actually meet user needs.
3Measurement precision
If emotion feature analysis is performed on historical stickers, then recommendation accuracy improves, but computational complexity increases
Solution Approach 1:
The patent extracts only the essential emotion features from historical sticker data rather than analyzing all possible attributes. By selectively extracting relevant emotion information (such as positive/negative sentiment, intensity levels), the system achieves good recommendation accuracy while keeping computational complexity manageable.
Solution Approach 2:
The system performs emotion feature analysis on a representative subset of historical stickers rather than every single sticker in a user's history. This partial analysis approach provides sufficient accuracy for personalization while significantly reducing the computational burden compared to comprehensive analysis.
Data Source
AI summary
Aspects of the disclosure provide methods and apparatuses for recommending a sticker set. An apparatus for recommending a sticker set includes interface circuitry and processing circuitry. When the interface circuitry receives a sticker recommendation request from a terminal, the processing circuitry determines a historical sticker set that includes a sticker previously sent by a user of the terminal device, and at least one recommendable sticker set not including the historical sticker set. Then the processing circuitry determines a recommendation index for each of the at least one recommendable sticker set according to an emotion feature of the historical sticker set and an emotion feature of the respective recommendable sticker set. According to the recommendation index for each of the at least one recommendable sticker set, the interface circuitry sends a sticker set recommendation for one or more of the at least one recommendable sticker set to the terminal device.