Conference Name Attribution Using Connection Strength Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional conference systems fail to accurately detect and attribute names, leading to transcription errors due to mispronunciations, nicknames, or multiple users with similar names, which complicates identification of the intended person during conferences.
Innovation Solution
A conference system that utilizes automated speech recognition (ASR) and machine learning (ML) to determine the intended person by processing real-time transcriptions, leveraging connection strength scores based on interaction history and recency, and contextual awareness to attribute names accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional speech recognition is used to transcribe conference audio, then transcription speed is maintained, but name detection accuracy deteriorates due to mispronunciations, nicknames, and similar names
Solution Approach 1:
The patent introduces an intermediary name resolution system that sits between speech recognition and the final transcript. This intermediary layer takes recognized names, compares them against contact lists, calculates connection strength scores based on interaction history, and resolves ambiguities before finalizing the transcript, thereby improving accuracy without sacrificing speed
Solution Approach 2:
The system implements feedback by continuously monitoring transcription quality and using connection strength metrics to refine name attribution. The feedback loop analyzes interaction patterns and adjusts name resolution decisions, reducing transcription errors while maintaining efficient processing
2Measurement precision
If connection strength scoring based on interaction history is implemented, then name attribution accuracy is improved, but system complexity increases
Solution Approach 1:
The system performs preliminary actions by pre-calculating and storing connection strength scores based on interaction history before name resolution is needed. Contact lists and interaction metrics are prepared in advance, allowing the name attribution process to quickly retrieve and use pre-computed data rather than analyzing all interactions in real-time
Solution Approach 2:
The patent changes parameters by using connection strength scores as a quantitative metric to resolve name ambiguities. Instead of relying on simple string matching, the system transforms the problem into a scoring mechanism that considers interaction frequency, recency, and context, improving accuracy while keeping the computational approach manageable
3Productivity
If real-time processing of conference audio is performed, then conference productivity is maintained, but computational resources are consumed
Solution Approach 1:
The system applies partial action by focusing computational resources only on name detection and resolution tasks rather than processing entire conference transcripts equally. It selectively analyzes segments containing potential names, using connection strength scoring only when necessary to resolve ambiguities, thereby reducing overall computational consumption while maintaining real-time productivity
Data Source
AI summary
A conference system detects a name of a person in a discussion between two or more participants in a conference. The name is detected in an audio component of the conference. The conference system determines a list of person candidates associated with the name based on scores determined for individual persons in a contact list relative to the name. The conference system determines a connection strength score between the conference participant that spoke the name and a person in the list of person candidates. The conference system determines that the person is an intended person based on the connection strength score and generates a transcript that includes the name of the intended person.