Question Embedding Categorization for Accurate Meeting Transcript Routing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to efficiently categorize questions from meeting transcripts and provide them to relevant targets with high accuracy while minimizing computational resources and configuration complexity.
Innovation Solution
A computerized method using a question detection model and a question clustering model with category vector embeddings to analyze transcripts, automatically categorize questions, and provide them to appropriate targets through a customizable interface.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional question categorization methods are used, then system simplicity is maintained, but categorization accuracy and efficiency deteriorate
Solution Approach 1:
The patent replaces traditional mechanical classification systems with vector embedding-based semantic representation. Questions and categories are transformed into vector spaces where semantic similarity can be computed efficiently, substituting rule-based mechanical categorization with mathematical vector operations that achieve higher accuracy without proportional increases in system complexity
Solution Approach 2:
The patent changes the parameter space by introducing vector embeddings as an intermediate representation layer. Instead of directly comparing categorical labels, the system transforms questions and categories into continuous vector spaces, enabling more nuanced similarity measurements and improving categorization precision through parameter transformation
2Measurement precision
If comprehensive question analysis is performed, then categorization accuracy improves, but computational resources increase
Solution Approach 1:
The patent performs preliminary action by pre-computing and storing category vector embeddings in a reference database before actual question categorization. This allows the system to avoid re-computing category representations for each question, significantly reducing computational resources during runtime while maintaining high categorization accuracy through pre-prepared semantic references
Solution Approach 2:
The patent extracts only the essential semantic features needed for categorization into compact vector representations, rather than processing entire question texts or performing comprehensive linguistic analysis. This extraction approach maintains categorization accuracy by preserving key semantic information while reducing computational burden through dimensionality reduction
3Productivity
If manual question routing is used, then system complexity is reduced, but time efficiency deteriorates
Solution Approach 1:
The patent implements self-service by enabling questions to automatically categorize and route themselves to appropriate targets without manual intervention. The vector embedding system allows questions to independently determine their categories and identify relevant recipients through semantic matching, achieving high routing efficiency while eliminating time wastage associated with manual question assignment
4Productivity
If vector embedding models are implemented, then categorization efficiency improves, but configuration complexity increases
Solution Approach 1:
The patent applies universality by using a general-purpose vector embedding framework that can handle multiple categories and question types through a unified mathematical approach. The same vector space and similarity computation mechanisms work across different domain categories, reducing configuration complexity compared to specialized classification systems that would require separate configurations for each category type
Data Source
AI summary
A computerized method categorizes questions from a transcript and provides those categorized questions to targets. A transcript associated with a meeting is obtained and a question in the obtained transcript is detected by a question detection model. A question vector embedding of the detected question is generated using a question clustering model and a category of the detected question is determined by the question clustering model using the generated question vector embedding and a plurality of category vector embeddings. A target of the detected question is identified using the determined category and the detected question is provided to the identified target via a question provision interface. Further, the question clustering model is used to generate the category vector embeddings from example questions of category question catalogs, such that the set of categories for which the question clustering model is configured can be efficiently modified.