Meeting Speech Extraction Using LLM-Defined Summary Items
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing information processing systems struggle to set customizable extraction items for generating minutes from communication content, resulting in uniform summaries despite varying meeting content.
Innovation Solution
An information processing system utilizing a large-scale language model (LLM) to analyze communication content, allowing customizable extraction items and generating business negotiation and analysis reports based on these items.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional speech recognition and summarization techniques are used, then meeting content can be processed and minutes can be generated, but extraction items cannot be customized and all meetings produce uniform summaries
Solution Approach 1:
The system dynamically adjusts extraction items based on user input and meeting context. The extraction items are not fixed but can be modified and customized for different meeting types and user needs, allowing the system to adapt its behavior rather than following a static configuration
Solution Approach 2:
The system allows changing parameters of the extraction process by accepting user-defined extraction items. Different meeting scenarios can specify different extraction parameters (such as focusing on decisions, action items, or specific topics), enabling flexible customization without redesigning the entire system
2Loss of information
If fixed extraction items are used for all meetings, then the system is simple to operate, but the extraction results are uniform and do not reflect specific meeting requirements
Solution Approach 1:
The system enables users to self-configure extraction items based on their specific meeting needs. Users can independently select and define what information to extract without requiring complex system administration or predefined templates, making the customization process simple and intuitive
Solution Approach 2:
The system maintains a universal framework that can handle multiple meeting types and extraction requirements through a single configurable interface. The same core processing engine serves diverse extraction needs by accepting different extraction item specifications, eliminating the need for separate systems for different meeting scenarios
3Measurement precision
If detailed customization of extraction items is enabled, then tailored summaries can be generated, but the operation complexity increases for users
Solution Approach 1:
The system performs preliminary processing of meeting content to identify potential extraction candidates before final extraction. This preliminary analysis prepares the data structure and identifies key elements, reducing the complexity of the actual extraction configuration while maintaining high precision results
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An information processing system (1) comprising: an acquisition unit (512) configured to acquire speech uttered by a first participant and a second participant, the first participant and the second participant participating in a communication session; a text data acquisition unit (513) configured to acquire text data from the speech through speech recognition; an operation receiving unit (523) configured to receive a setting of an extraction item to be extracted from the text data; and an extraction result acquisition unit (514) configured to acquire an extraction result obtained by extracting first utterance information related to the extraction item from the text data.