Iterative Video Recommendation Query Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for identifying related videos on video hosting websites often fail to present content that is truly of interest to users, relying on keywords or popularity, which can lead to irrelevant recommendations.
Innovation Solution
A video hosting website generates an ordered list of words from the metadata of a selected video, modifies search queries by adding or removing words to achieve a desired number of search results within a set range, and processes these queries to provide a set of related videos that are dynamically adjusted based on user preferences and metadata analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional methods use keywords or popularity to identify related videos, then the process is simple and fast, but the relevance and user interest of recommended videos deteriorates
Solution Approach 1:
The system dynamically adjusts search queries by iteratively adding or removing words based on the number of search results obtained, transforming a static keyword search into an adaptive process that responds to feedback from search result counts
Solution Approach 2:
The system uses feedback from search result counts to modify subsequent search queries, creating a closed-loop process where the number of results directly influences the composition of the next query through iterative addition or removal of words
2Measurement precision
If the search query includes more words from metadata, then the precision of video matching improves, but the number of search results decreases
Solution Approach 1:
The system dynamically balances query precision and result quantity by iteratively adjusting the number of words in the search query based on whether the result count falls within the desired range
Solution Approach 2:
The system changes the parameter of query word count to achieve the desired balance between matching precision and result quantity, adjusting this parameter iteratively until results fall within the specified range
Data Source
AI summary
The present invention provides methods and computer program products for identifying a set of related videos for a first video. An embodiment of the method comprises a video hosting website creating an ordered list containing a plurality of words from the metadata associated with the first video. The video hosting website uses a select number of words from the ordered list to form a search query. The search query is processed by the videos hosting website to obtain search results that contain a number of related videos. If the number of related videos in the search results is not within a set range, the search query is modified and the modified search query is processed. If the number of related videos in the search results is within a set range, the search results are provided as the set of related videos for the first video.