Content Recommendation Control Using Temporal User Viewing Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing content recommendation systems struggle to account for varying user preferences over time and fail to recommend diverse content due to reliance on static viewing history and lack of time-based analysis.
Innovation Solution
An electronic apparatus utilizing a Graph Neural Network (GNN) model processes user content viewing history to group users into viewing groups based on time sequences and attributes, providing personalized recommendations and joint viewing opportunities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If only viewing history of one user is considered in determining recommended content, then the system is simple to operate, but various types of content cannot be recommended
Solution Approach 1:
The patent segments users into multiple viewing groups based on their viewing history patterns. Each group represents users with similar preferences, allowing the system to recommend diverse content types by targeting different segments rather than treating all users uniformly.
Solution Approach 2:
The patent introduces a new dimension by creating graph information that represents viewing history in a structured format with nodes and edges. This graph structure enables multi-dimensional analysis of user preferences, allowing the system to discover diverse content patterns that would be invisible in traditional flat analysis.
2Adaptability or versatility
If time information is not reflected in determining recommended content, then the system is simpler, but preferences that vary depending on viewing time cannot be reflected
Solution Approach 1:
The patent makes the recommendation system dynamic by incorporating time information into the graph structure. Viewing history is represented with temporal relationships, allowing the system to adapt recommendations based on when users view content, capturing time-varying preferences.
Solution Approach 2:
The patent performs preliminary action by pre-processing viewing history into graph information that includes temporal attributes. This preparation enables the system to efficiently query time-based patterns during recommendation generation without adding complex real-time analysis.
3Adaptability or versatility
If the number of content varies in the operation of determining recommended content, then more content options are available, but it becomes difficult to recommend content not reflected in viewing history
Solution Approach 1:
The patent introduces graph information as an intermediary representation between raw viewing history and recommendation output. This graph structure acts as a mediator that organizes diverse content options while maintaining precise relationships to user preferences, enabling both broad coverage and accurate matching.
Solution Approach 2:
The patent replaces traditional mechanical content matching systems with a graph-based neural network approach. This substitution enables the system to handle large numbers of content items while maintaining recommendation precision through learned patterns in the graph structure rather than rule-based matching.
Data Source
AI summary
An electronic apparatus is provided. The electronic apparatus includes a memory configured to store at least one command and a processor connected to the memory and configured to control the electronic apparatus. The processor is configured to, by executing the at least one command, obtain graph information for a plurality of users in which viewing content is arranged in a time sequence based on content viewing history information for a plurality of users, obtain attribute information for each content by inputting the graph information for the plurality of users to a Graph Neural Network (GNN) model, group the plurality of users into a plurality of viewing groups based on the graph information for the plurality of users and the attribute information for each content, and provide a recommended content based on content information corresponding to the plurality of viewing groups.