Media Recommendation Engine Cold Start via External Provider Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In media streaming and online content contexts, new subscribers often lack personalized content recommendations due to unknown user preferences and demographics, and existing content recommendation engines function in isolation, leading to inefficiencies and user frustration, especially when recommended content is unavailable from the content library or schedule.
Innovation Solution
A content recommendation engine that queries partner service providers for user interaction histories to create personalized profiles and recommends content based on user preferences, and arranges for unavailable content to be made available through partnerships with other providers, ensuring timely and relevant content delivery.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a content recommendation engine is used to provide personalized recommendations, then user satisfaction and content delivery effectiveness are improved, but the engine cannot provide accurate recommendations for new subscribers due to lack of user preference data (cold start problem)
Solution Approach 1:
The system performs preliminary actions by proactively reaching out to external providers before a new subscriber can provide any preferences. It automatically queries multiple content providers about the user's subscription status and retrieves interaction histories, watchlists, and profile data in advance, so that personalized recommendations can be generated immediately upon subscription without requiring user input during the cold start phase.
Solution Approach 2:
The patent introduces an intermediary component that acts as a bridge between the content recommendation engine and external content providers. This intermediary systematically queries external providers for user data, aggregates information from multiple sources, and supplies the recommendation engine with comprehensive user profiles, thereby resolving the information gap during cold start without requiring direct user interaction.
2Measurement precision
If content recommendation engines function independently in isolation, then system simplicity is maintained, but the ability to provide comprehensive and accurate recommendations is limited
Solution Approach 1:
The recommendation system is designed with multi-functionality by integrating capabilities to query multiple external content providers, aggregate data from diverse sources, and generate recommendations across different content types. The system serves multiple purposes: retrieving user profiles, checking subscription statuses, gathering interaction histories, and providing recommendations, all within a unified architecture that manages complexity through standardized interfaces.
Solution Approach 2:
An intermediary layer is introduced to manage the complexity of interacting with multiple external providers. This intermediary handles data aggregation, normalization, and coordination between different sources, allowing the core recommendation engine to function independently while still accessing comprehensive data through the intermediary's standardized data access layer.
3Measurement precision
If the recommendation engine recommends content that is not available in the content library or schedule, then user preferences are accurately reflected, but user frustration increases due to inability to access recommended content
Solution Approach 1:
The intermediary component queries external content providers to verify the availability of recommended content. When the recommendation engine identifies content based on user preferences, the intermediary checks with external providers to confirm whether that content is currently available in their libraries or schedules, and only presents recommendations for content that can actually be accessed by the user.
Solution Approach 2:
The system implements feedback mechanisms where the availability status of recommended content is continuously monitored and verified. If recommended content becomes unavailable, the system receives feedback about this unavailability and adjusts its recommendations accordingly, presenting alternative content that is both aligned with user preferences and currently accessible through partnered providers.
Data Source
AI summary
Media content recommendations and user profile display recommendations, including user names and icons of users associated with the subscription, may be provided, for example, upon cold start of a user subscription. A content provider may request a history of user interactions and profile information from a second content provider and then determine that the content items were not previously consumed by this user, and transmit content recommendations accordingly. Generating for play content items not available from a first content provider but available from a second content provider is also contemplated. The content item may be transmitted to the user device via an application associated with the first content provider on the user device, or via an application associated with the second content provider on the user device.