Profile-Based Channel Recommendations for Streaming Services
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recommender systems in streaming TV services face a 'cold start' dilemma, where new users without sufficient watch history cannot receive personalized content recommendations.
Innovation Solution
The method involves identifying existing users similar to new users based on their profiles, using a knowledge graph to calculate similarity scores, and recommending channels that are popular among these similar users.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If collaborative filtering is used to generate personalized recommendations, then recommendation accuracy is improved, but new users without watch history cannot receive meaningful recommendations
Solution Approach 1:
The system performs preliminary actions by pre-calculating and storing user profile features and channel preference models before new users arrive. When a new user signs up, the system immediately queries pre-computed profile features and matching channel preferences from the database, enabling instant recommendations without requiring the user to have any watch history yet.
Solution Approach 2:
The patent introduces an intermediary mechanism using a database that stores pre-computed user profile features and channel preference models. This intermediary layer allows the system to match new users to preferred channels based on their profile attributes (age, location, device) without needing direct interaction history, thus bridging the cold start problem.
2Adaptability or versatility
If profile-based recommendations are used for new users, then new user onboarding is improved, but system complexity increases due to knowledge graph construction
Solution Approach 1:
The system segments the recommendation task into two independent parts: (1) building a knowledge graph with user profiles and channel preferences, and (2) querying this graph for recommendations. By separating the complex knowledge graph construction from the simple query operation, the system can handle new user onboarding without requiring real-time complex computations during user signup.
Solution Approach 2:
The knowledge graph and user profile features are pre-computed and stored in the database before new users arrive. When a new user signs up, the system only needs to query pre-computed profile features and match them against stored channel preference models, avoiding the complexity of building the knowledge graph in real-time.
3Ease of operation
If real-time recommendation generation is implemented, then user experience is improved, but processing speed decreases due to cold start computations
Solution Approach 1:
The system performs all complex matching computations in advance by pre-computing user profile features and channel preference models and storing them in the database. When a new user arrives, the system only executes simple queries against the pre-computed data, achieving both fast processing and good user experience without cold start delays.
Data Source
AI summary
Described herein are method, systems, and media for recommending channels to a new user in a streaming service. In an embodiment, an exemplary method includes identifying, from a plurality of existing users of the streaming service, one or more existing users that each are similar to a new user of the streaming service based on their respective profiles; identifying one or more channels recommended to each of the one or more identified existing users; and displaying, on a streaming device of the new user, at least one channel of combined channels of the one or more channels recommended to each of the one or more identified existing users.