Sorted List Generation Using Clustering and Credit-Based Scheduling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems fail to effectively integrate the lean-back TV viewing experience with internet video browsing, as users must actively search for videos on platforms like YouTube, lacking personalized and diverse results.
Innovation Solution
An electronic device and method that generates queries based on TV program metadata to find related internet videos, using clustering and credit-based scheduling to balance relatedness and diversity in search results, ensuring the most relevant and diverse items appear at the top of the list.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If internet videos are searched using simple keyword queries, then search speed is improved, but result relevance and personalization deteriorate
Solution Approach 1:
The patent introduces an intermediary system that sits between the user's simple keyword input and the video repository search. This intermediary automatically generates multiple refined queries based on the seed TV program metadata, acts as a mediator that transforms simple keywords into sophisticated search queries, thereby maintaining search speed while improving result relevance through automated query expansion and refinement
Solution Approach 2:
The system performs preliminary actions by pre-generating multiple search queries based on the seed TV program's metadata before the actual search execution. The queries are prepared in advance with different angles and perspectives, so when the search is initiated, these pre-prepared queries are executed simultaneously or sequentially, achieving both speed through preparation and relevance through comprehensive query coverage
2Adaptability or versatility
If multiple queries are generated to improve result diversity, then result diversity is improved, but system complexity deteriorates
Solution Approach 1:
The patent segments the search process into distinct components: the seed item processor that generates multiple queries from the TV program metadata, the query dispatcher that manages multiple search operations, and the result aggregator that combines and ranks results. This segmentation allows each component to handle a specific aspect of the complexity, making the overall system more manageable while achieving diverse results through coordinated multiple queries
Solution Approach 2:
The system implements a universal query generation mechanism that can handle different types of TV programs (movies, series, documentaries) and generate appropriate search queries automatically. The same core infrastructure handles diverse content types by adapting query generation based on the seed item's metadata, achieving result diversity across different content categories without requiring separate specialized systems for each type
3Measurement precision
If results are sorted strictly by relatedness to seed item, then result relevance is improved, but result diversity deteriorates
Solution Approach 1:
The patent applies local quality by differentiating the treatment of results from different query sources. Results from queries that explore different aspects of the seed item (different genres, different media types, different creators) are given different weights and consideration in the final ranking. This allows the system to maintain high relevance for each query's results while ensuring diversity in the overall result set by locally adjusting ranking criteria based on the query's purpose and the characteristics of results from different sources
Data Source
AI summary
An electronic device and method for automatic generation of a sorted list of items related to a seed item comprises a relatedness determinator to compare the seed item with a plurality of further items and to determine a relatedness value for each further item with respect to the seed item. The device also has a clustering engine to cluster the further items by determining a relative relatedness between (among) the further items. Each further item is assigned to one cluster. The device also has a list generator to generate a sorted result list by sorting the further items according to both, their relatedness value and their belonging (or membership) to a cluster in that once an item is added to the sorted list, the relatedness value depending ranking of the further items in that cluster is at least momentary lowered so as to promote adding items of further clusters.