Knowledge Graph Content Recommendations Using Fuzzy Logic
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional content recommendation systems face challenges in providing personalized recommendations to new subscribers due to the lack of viewing history data, relying on basic generic metadata, and failing to account for user-specific preferences and interests effectively.
Innovation Solution
A method that accesses a knowledge graph based on content attributes, user preferences, and metadata to generate personalized content recommendations, using fuzzy logic and user interface adjustments to prioritize attributes of interest, and incorporates feedback from content trailers to refine recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional recommendation systems use basic generic metadata and collaborative filtering, then they can provide recommendations to all users, but they fail to provide personalized recommendations to new subscribers due to lack of viewing history data
Solution Approach 1:
The system performs preliminary actions by collecting user feedback on content trailers during the sign-up phase before actual content consumption begins. This advance data collection enables the knowledge graph to be populated with user preferences early, eliminating the cold start problem for new subscribers and allowing personalized recommendations from day one rather than waiting for viewing history to accumulate
Solution Approach 2:
The patent introduces a knowledge graph as an intermediary structure that connects content attributes with user preferences. This knowledge graph serves as a mediator between the content catalog and recommendation engine, enabling personalized recommendations by translating generic metadata into user-specific recommendations through structured relationships and fuzzy logic matching
2Productivity
If conventional systems rely on viewing history and collaborative filtering, then they can leverage existing user data, but they cannot effectively recommend content to new subscribers without established viewing patterns
Solution Approach 1:
The system performs preliminary data collection during the sign-up phase by presenting content trailers and capturing user feedback before actual content consumption. This advance collection of preference data eliminates the need to wait for viewing history to accumulate, enabling immediate personalized recommendations without the typical cold start delay
Solution Approach 2:
The system implements a feedback mechanism where user responses to content trailers during sign-up are immediately processed to update the knowledge graph and refine recommendations. This real-time feedback loop allows the system to learn user preferences quickly and adjust recommendations dynamically, achieving high accuracy without requiring extensive historical data
3Adaptability or versatility
If conventional recommendation systems use simplistic attribute matching, then they are easy to implement, but they fail to account for user-specific preferences and interests effectively
Solution Approach 1:
The knowledge graph acts as an intermediary layer between simple content metadata and complex recommendation logic. It structures content attributes and user preferences into standardized relationships, enabling sophisticated personalized recommendations while maintaining a manageable system architecture that doesn't require complex algorithms at every level
Solution Approach 2:
The system employs fuzzy logic to handle parameter matching between content attributes and user preferences. Instead of rigid binary matching, fuzzy logic allows for degree-based matching that captures nuanced user preferences, enabling more adaptable recommendations while working within the existing metadata structure rather than requiring complete system redesign
Data Source
AI summary
Improved content recommendations are generated based on a knowledge graph of a content item, which is based on an attribute of the content item, metadata regarding the content item, a viewing history, and user preferences determined by analysis and selected by a user. An option for selecting attributes of interest from a plurality of attributes is generated for display. A content recommendation based on the selected attributes is generated and displayed in a user interface, which changes as user preference selections change. As a result, a user quickly identifies and consumes a customized list of content items related to the user's favorite actor, character, title, depicted object, depicted setting, actual setting, type of action, type of interaction, genre, release date, release decade, director, MPAA rating, critical rating, plot origin point, plot end point, and the like. Related apparatuses, devices, techniques, and articles are also described.