Video Content Identification Using Bibliographic Occurrence Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Video search engines often produce imprecise results due to the abundance of video content with thin metadata, leading to unrelated videos being matched to a user's query.
Innovation Solution
A method that formulates a query based on terms from a video bibliographic entry, calculates occurrence scores for search results based on the presence and relevance of these terms within associated text, and selects resources for storage, improving the identification of matching video content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If video search engines rely on metadata and textual descriptions for video identification, then the search process is simple and fast, but the precision of video content identification deteriorates due to thin metadata and abundant video content from various sources
Solution Approach 1:
The patent segments the video identification process into multiple independent components: extracting video content features (audio, visual, text), extracting search query features, computing similarity scores for each feature type, and aggregating these scores. This segmentation allows each component to be processed independently and efficiently while collectively achieving high identification accuracy.
Solution Approach 2:
The patent introduces an intermediary similarity scoring mechanism that bridges the gap between simple metadata matching and complex video content analysis. By computing similarity scores based on multiple feature types (audio, visual, textual) and aggregating them, the system achieves accurate video content identification without requiring full video analysis, thus maintaining search efficiency while improving precision.
2Adaptability or versatility
If video search engines include multiple matching videos in search results, then more potential matches are provided, but the relevance assessment becomes more difficult and user satisfaction decreases due to unrelated content
Solution Approach 1:
The patent implements a feedback mechanism where similarity scores are computed for each video against the search query, and these scores are used to rank and filter results. The system provides feedback by highlighting the most relevant videos at the top of results and suppressing unrelated content, enabling users to quickly assess relevance without manually evaluating multiple potentially unrelated videos.
Solution Approach 2:
The patent changes the parameter of result presentation by transforming raw similarity scores into ranked positions and applying threshold-based filtering. This parameter transformation allows the system to maintain adaptability by including multiple matching videos while improving precision by ranking them according to relevance, thus presenting only the most relevant results to users.
Data Source
AI summary
The present disclosure relates to the identification of video content. In one aspect, a method includes generating a query based on bibliographic data. The method also includes obtaining a collection of resources responsive to the query, wherein one or more of the resources include text and video content. The method further includes calculating occurrence scores for the resources. A particular occurrence score for a particular resource is based at least in part on the bibliographic data matching text included in the particular resource and the text being associated with video content. The method further includes selecting one or more resources as including video content identified by the bibliographic data using the occurrence scores. The method further includes storing data associating the selected resources with the bibliographic data.