Implicit Recommender Negative Example Intermediary
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Television recommender systems face challenges in generating effective negative examples for user profiles, as they typically rely on positive viewing history data, lacking sufficient negative data points to provide accurate recommendations.
Innovation Solution
The method involves receiving and determining recommendations based on both negative and positive examples from other users, selecting users based on geographical location or similarity in likes and dislikes, and mapping these examples to an electronic program guide database to generate user-specific recommendations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If the system uses only positive viewing history data to generate recommendations, then the system can operate with simple data collection, but the recommendation accuracy deteriorates due to insufficient negative data points
Solution Approach 1:
The patent introduces other users as intermediaries to provide negative examples. Instead of requiring the target user to explicitly provide negative feedback, the system collects negative examples from other users' viewing histories and uses these as mediators to train the recommendation model, thereby improving accuracy without complicating the target user's data collection burden
Solution Approach 2:
The patent copies negative example data from other users' viewing histories to train the recommendation model. By replicating negative feedback data from multiple users, the system creates sufficient negative data points without requiring each user to explicitly provide their own negative examples, thus maintaining data collection simplicity while improving model accuracy
2Measurement precision
If the system collects negative examples from multiple users, then the recommendation accuracy improves, but the system complexity increases due to user selection and data mapping requirements
Solution Approach 1:
The patent segments the data collection process into distinct modules: user profile generation, negative example collection from other users, example mapping to EPG database, and recommendation generation. This segmentation allows each component to be developed and optimized independently, reducing overall system complexity while maintaining high accuracy
Solution Approach 2:
The patent creates a universal data mapping mechanism that works across different users and EPG databases. The system uses a general approach to map negative examples from various users to the target user's EPG database, making the system adaptable to different contexts without requiring user-specific complex processing for each case
Data Source
AI summary
A method for generating a recommendation is provided. The method includes the steps of: receiving at least one of a negative and positive example from one or more other users; and determining a recommendation for a user based on at least one of the negative and positive examples. The method can further include the step of generating a user profile for the user based on previous behavior of the user, in which case the determining step can include determining the recommendation based on the negative example and the user profile. The determining step can also include the step of determining the recommendation based on both of the negative and positive examples.