LLM Rule Extraction for Faster Transaction Exception Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current techniques for evaluating transactions using large language models (LLMs) often require numerous calls, leading to reduced computer efficiency and delays due to resource tying.
Innovation Solution
A system that parses documents to identify conditional language, generates rules from the language using an LLM, associates these rules with document sections, and executes them against transactions to discover exceptions, generating a narrative explanation for violations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the system parses contracts on the fly and asks the LLM what parts apply to each transaction, then the system can identify applicable contract terms, but this results in many added calls to the LLM which reduces computer efficiency and causes delays
Solution Approach 1:
The system extracts and stores contract terms and conditions in a structured format during an initial parsing phase, before transactions are evaluated. This preliminary structuring of contract data eliminates the need for repeated LLM calls during transaction evaluation, as the system can directly query the pre-processed contract structure.
Solution Approach 2:
The patent introduces an intermediary structured data layer between the contract document and the transaction evaluation process. This intermediary structure acts as a mediator that translates contract language into machine-queryable formats, reducing direct dependence on LLM calls for each evaluation.
2Loss of information
If the system asks the LLM what parts of the contract apply to each transaction, then the system can provide explanations for exceptions, but this ties up resources and results in delays
Solution Approach 1:
The system pre-processes contract documents to extract and structure relevant terms, conditions, and their interrelationships before evaluation. This preliminary action creates a ready-to-query structure that enables rapid exception explanation without requiring real-time LLM analysis for each transaction.
Solution Approach 2:
The patent creates structured copies of contract information in an optimized format that preserves the semantic meaning needed for explanations. These structured copies enable the system to generate explanations efficiently by querying the pre-processed data rather than repeatedly invoking the LLM.
Data Source
AI summary
A computer-implemented method for evaluating transactions against rules includes parsing sections of a document with conditional language and generating rules from the conditional language using a large language model. The rules are associated with corresponding sections of the document such that when a rule fires the rule is associated with the corresponding sections of the document. The rules are executed against transactions to discover exceptions. A narrative is generated to explain the exceptions to a user. The narrative includes the rules with the corresponding sections and an explanation.