Search-Driven Video Program Extracts from Subtitle-Matched Scenes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Identifying interesting scenes in video programs for viewers with limited time is time-consuming and varies based on individual evaluation, and existing methods like US2012150907A1 do not effectively generate concise and representative video extracts.
Innovation Solution
Correlate user search queries to video program subtitles using voice recognition and matching techniques to identify scenes of interest, stitching together relevant portions to create video extracts that reflect user interest, and optionally using social media data to enhance relevance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual evaluation of video programs is used to identify interesting scenes, then the quality of scene selection can be maintained, but the time consumption and labor requirements increase significantly
Solution Approach 1:
The patent replaces manual mechanical evaluation with automated systems including voice recognition software to transcribe and analyze video content, search query processing systems to identify user interests, and automated stitching software to compile highlights. This substitution maintains scene selection quality while eliminating time-consuming manual labor.
Solution Approach 2:
The system enables self-service by using user-generated search queries as the basis for automatic scene identification. The system processes user interests directly without requiring manual curation, automatically generating personalized video extracts based on what users are already searching for.
2Productivity
If automated scene detection is used to reduce time consumption, then the speed of extract generation improves, but the accuracy and representativeness of selected scenes may deteriorate
Solution Approach 1:
The patent introduces search queries as an intermediary between automated detection and scene selection. Instead of directly analyzing video content alone, the system uses user search queries to guide which automated detections are relevant, ensuring both speed and accuracy by filtering through user interest signals.
Solution Approach 2:
The system incorporates feedback loops where search query results inform subsequent scene selections. The automated system continuously refines its scene identification based on what users are actually searching for, improving accuracy over time while maintaining high generation speeds through automated processing.
3Loss of information
If video extracts include all matched scenes to ensure completeness, then the representativeness of the extract improves, but the length and conciseness deteriorate
Solution Approach 1:
The patent extracts only the most relevant portions of matched scenes rather than including all of them. By selectively taking out key segments that best represent user interests, the system maintains content representativeness while keeping the final extract concise and manageable in length.
Solution Approach 2:
The system segments video content into discrete searchable units and selectively assembles only those segments that match user queries. This segmentation approach allows the extract to include sufficient representative content while avoiding unnecessary length by excluding non-matching portions.
4Adaptability or versatility
If personalized video extracts are generated for individual users, then the relevance to user interests improves, but the complexity of the system increases
Solution Approach 1:
The patent implements a universal search query processing framework that serves all users through the same system infrastructure. The same automated detection, voice recognition, and stitching systems handle personalized requests, achieving personalization without proportionally increasing overall system complexity through shared components.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A process builds extracts for video programs based on search query spikes during transmission of the video programs. The process is performed at a server system with one or more processors and memory storing programs configured for execution by the one or more processors. The process identifies a plurality of search query spikes from search queries submitted by a plurality of users, and correlates a subset of the search query spikes to a broadcast video program. Each correlated search query spike corresponds to a respective location in the video program. The process constructs a snippet of the video program that contain the locations corresponding to the correlated search query spikes. In some implementations the portions extend to video scene boundaries before and after each location. In some implementations, video program subtitles are used to match search query spikes to video program locations.