Historical Data Tacit Knowledge Extraction With LLM Hyperspaces
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems fail to effectively extract tacit knowledge from historical data, particularly from unstructured sources, and lose this knowledge over time due to lack of documentation, and existing techniques are inefficient in retaining domain flavor and business rules.
Innovation Solution
A method and system that utilize pre-trained large language models (LLMs) to convert large dimensional hyperspaces of tacit knowledge into smaller dimensional hyperspaces by selecting effective knowledge points, iteratively refining domain rules and LLMs to achieve desired output quality, and store the refined models for future use.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If all knowledge points from historical data are retained in the hyperspace, then the completeness of tacit knowledge is improved, but the processing time and computational resources increase significantly
Solution Approach 1:
The system extracts only the essential and relevant knowledge points from the historical data to form a reduced hyperspace. The LLM identifies and selects critical knowledge points that are most relevant to downstream tasks, discarding redundant information. This extraction process maintains the essential tacit knowledge while significantly reducing processing requirements.
Solution Approach 2:
The knowledge space is segmented into different dimensions and layers. The system creates a hierarchical structure where knowledge points are organized from general to specific, allowing selective processing of different segments based on task requirements. This segmentation enables efficient retrieval and processing of only the necessary knowledge portions.
2Measurement precision
If domain rules are continuously refined through iterative feedback, then the accuracy of downstream task performance is improved, but the complexity of the system increases
Solution Approach 1:
The system implements an iterative feedback mechanism where the performance of downstream tasks is continuously evaluated. Based on performance metrics, the system automatically refines domain rules and re-trains the LLM. This closed-loop feedback process gradually improves accuracy without requiring manual intervention, managing complexity through automation.
Solution Approach 2:
The system performs preliminary extraction and organization of domain rules from historical data before the iterative refinement process begins. This preliminary structuring of knowledge and rules simplifies subsequent iterations by providing a solid foundation, reducing the complexity burden during the refinement phases.
3Reliability
If pre-trained LLMs are fine-tuned with updated domain rules, then the quality of extracted tacit knowledge is improved, but the computational resources and training time increase
Solution Approach 1:
Instead of complete re-training of the LLM with all possible data, the system applies partial fine-tuning using only the updated domain rules and relevant feedback from iterative processes. This partial action approach achieves the necessary quality improvement in tacit knowledge extraction without the excessive computational cost of full re-training, optimizing the balance between quality and resource consumption.
Data Source
AI summary
Business rules are currently not documented and are present only as knowledge with subject matter experts (SMEs). The knowledge can be lost with time if it is not extracted or recorded. Existing techniques are unable to extract tacit knowledge and to retain the domain flavor in extracted information. Present disclosure provides a method and a system for extracting tacit knowledge from historical data. The system represents each point in historical data as a large dimensional hyperspace which contains all unstructured information where tacit knowledge can exist. Then, system maps large dimensional hyperspace to smaller dimensional hyperspace using pre-trained large language model (LLM). Thereafter, system, based on the series of downstream tasks, generates a feedback loop to optimally compute dimension of the smaller dimensional hyperspace. Once reduced dimensional space containing effective tacit knowledge information is available, system performs a downstream task based on the extracted tacit knowledge using another pre-trained LLM.