AI Task Routing for Condition-Specific Meeting Records
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video conferencing-based services struggle to accurately identify user intentions and provide meeting records that conform to specific user conditions, such as summaries for specific speakers or output in a set tone, often resulting in inaccurate results.
Innovation Solution
A method and system utilizing an artificial intelligence model to determine user tasks, such as summarization, question answering, or translation, by splitting meeting materials into units, performing tasks in parallel, and merging results to meet user requirements, including specific-speaker, action item, and condition-based summarization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If artificial intelligence models are used to generate meeting records with specific conditions (e.g., summaries for specific speakers, specific tones), then service customization and accuracy should improve, but in typical services users do not often receive accurate results according to the conditions they requested
Solution Approach 1:
The patent segments the meeting record generation process into multiple independent modules: a task determination device that identifies user intentions, a summarization task device that processes specific conditions, and a post-task device that refines output. This segmentation allows each module to specialize in specific conditions, improving overall accuracy and reliability of condition compliance.
Solution Approach 2:
The patent implements a feedback mechanism where the task determination device continuously monitors user inputs and adjusts task assignments accordingly. When users specify conditions like specific speakers or tones, the system receives feedback about these requirements and modifies the summarization process to ensure accurate compliance, thereby improving both measurement precision and reliability.
2Adaptability or versatility
If multiple types of summarization tasks are performed to meet diverse user requirements (specific speaker, action item, condition-based), then service versatility improves, but processing time and system complexity increase
Solution Approach 1:
The patent performs preliminary action by pre-defining multiple types of summarization tasks (specific speaker summarization, action item summarization, condition-based summarization) and their corresponding processing logic before user requests arrive. When users make diverse requirements, the system can quickly match and execute pre-configured task types without time-consuming on-the-fly decisions, thus maintaining versatility while reducing processing time.
Solution Approach 2:
The patent applies partial action by selectively executing only the necessary summarization tasks based on user requirements. Instead of always performing all possible summarization types, the system determines which specific tasks are needed (e.g., only specific speaker summarization when that condition is requested) and executes only those, thereby reducing unnecessary processing time while maintaining full versatility when needed.
3Productivity
If meeting materials are split into units for processing, then processing efficiency and model compatibility improve, but the complexity of coordinating multiple unit tasks increases
Solution Approach 1:
The patent segments meeting materials into units to improve processing efficiency and model compatibility. Each unit can be processed independently by the summarization task device, allowing parallel processing and better alignment with model capabilities. The segmentation is managed through a coordinated framework that tracks units through the summarization pipeline, reducing overall processing time while maintaining manageable coordination complexity.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A method, apparatus, system, and non-transitory computer program for providing a meeting record service is provided. The method includes determining a first task that a user wishes to perform from among multiple types of predetermined tasks, based on a user input, by implementing an artificial intelligence model in a task determination device, performing the determined first task, based on the meeting record, by implementing the artificial intelligence model in a first task device corresponding to the first task, and providing a result of performing the determined first task to the user.