Subtitle-Guided Semantic Video Search for Unclear Queries
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video search engines often return inaccurate results when users input unclear search queries, as they rely solely on title and tag comparisons, failing to capture the user's intended content.
Innovation Solution
A method utilizing a server that processes video files with subtitle analysis to generate summaries and tags, employs word embeddings and co-occurrence data to refine search queries, and presents target video segments based on candidate tag words selected by the user.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If video search engines compare search queries with video titles and tags, then the search process is simple and fast, but the search accuracy deteriorates when users input unclear search queries
Solution Approach 1:
The patent introduces subtitle text as an intermediary layer between the user's search query and the video content. The search engine first matches the query against subtitle text to identify candidate video segments, then presents relevant portions to the user. This intermediary approach enables more accurate semantic matching compared to direct title/tag comparison, thereby improving search accuracy while maintaining reasonable processing speed.
Solution Approach 2:
The patent segments the video content by generating separate subtitle files for different video segments. Instead of treating each video as a single unit with one title and tag set, the system divides videos into multiple segments with individual subtitles, allowing precise localization and matching of specific content portions to search queries, thus improving both accuracy and relevance.
2Device complexity
If video search engines rely solely on title and tag comparisons, then the system complexity is low, but the ability to capture user's intended content deteriorates
Solution Approach 1:
The patent performs preliminary processing of video content by generating subtitle files and extracting key information from subtitles before the search operation. This pre-processing step creates a rich textual representation of video content that captures the actual spoken information, enabling more accurate search matching without requiring complex real-time analysis during the search process itself.
Solution Approach 2:
The patent replaces the traditional mechanical comparison of titles and tags with a semantic search approach based on subtitle text analysis. Instead of simple string matching against metadata, the system uses the subtitle content—which reflects the actual spoken information—to understand and match user intent, thereby reducing information loss and improving content understanding accuracy.
Data Source
AI summary
A method for video searching is implemented using a server that stores a plurality of video files. The method includes: obtaining a searching word embedding that represents a search query received from a user device; obtaining a reference word embedding and a reference tag word based on the searching word embedding; obtaining, based on the reference tag word and a plurality of co-occurrences stored in the server, a number of associated co-occurrences, the reference tag word and the number of associated co-occurrences serving as candidate tag words; obtaining a target word embedding that corresponds with one of the candidate tag words, and selecting, based on the target word embedding, at least one target tag word embedding; and selecting, based on the at least one target tag word embedding, at least one target video segment, and presenting the at least one target video segment to the user device.