Transformer Contract Clause Extraction for Scalable Risk Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional contract risk assessment systems face scalability issues due to reliance on human-designed rules and require extensive re-coding for different contracts, and ML-based approaches are resource-intensive for comprehensive semantic analysis.
Innovation Solution
A computer-implemented contract risk assessment system leveraging transformer-based pre-trained large models (PLMs) to identify and highlight risk-prone clauses using a question-answer paradigm, employing transfer learning with a dataset like CUAD to enhance performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional rule-based systems are used for contract risk assessment, then comprehensiveness of analysis is improved, but device complexity and scalability deteriorate due to extensive re-coding requirements for different contracts
Solution Approach 1:
The patent replaces manual rule-based analysis with an automated machine learning system that uses natural language processing to identify and assess risk clauses. The system automatically processes contract text through trained models to detect risk patterns without requiring manual rule configuration, thereby reducing system complexity while maintaining comprehensive analysis capabilities.
Solution Approach 2:
The system changes the approach from fixed rule-based parameters to dynamic machine learning parameters that can adapt to different contract types. By using trained models that learn from data rather than relying on predetermined rules, the system achieves scalability across different contracts without extensive re-coding.
2Measurement precision
If ML-based approaches are used for comprehensive semantic analysis, then measurement precision is improved, but use of energy and computational resources worsen
Solution Approach 1:
The system extracts and focuses only on the most critical risk-related segments of contract text rather than performing exhaustive analysis on the entire document. By identifying and isolating key clauses that contain risk indicators, the system reduces computational resources required while maintaining high semantic analysis accuracy for the most important portions.
Solution Approach 2:
The system performs partial analysis by concentrating computational effort on specific high-risk sections of contracts identified through preliminary processing. Rather than analyzing every portion of the contract with equal depth, the system applies intensive semantic analysis only where risk indicators are present, reducing overall resource consumption while maintaining precision where it matters most.
3Reliability
If traditional systems analyze entire contracts, then completeness of assessment is improved, but loss of time and processing efficiency worsen
Solution Approach 1:
The system segments contracts into distinct clauses and sections, then applies risk assessment specifically to relevant segments rather than processing the entire document uniformly. This segmentation allows the system to maintain complete assessment of risk-prone areas while significantly reducing processing time by avoiding unnecessary analysis of low-risk portions.
Solution Approach 2:
The system extracts and prioritizes risk-prone clauses from the full contract text for detailed analysis. By identifying and isolating segments that contain potential risks based on preliminary processing and keyword detection, the system ensures complete assessment of critical areas while reducing overall processing time by minimizing analysis of non-critical portions.
Data Source
AI summary
Methods, systems, and computer-readable storage media for receiving a document provided as a computer-readable file, receiving a set of questions, for each question in the set of questions, generating an inference input including a question, at least a portion of text of the document, and multiple tokens, processing, by a PLM, the inference input to generate a set of text embeddings, processing, by a neural network, the set of text embeddings to provide sets of tokens, each set of tokens being specific to a segment of the document and including a start token and an end token respectively identifying a start position and an end position of the segment, determining, from the sets of tokens, a segment for display, and displaying at least a portion of the document in a UI and an annotation indicating the segment within the at least a portion of the document.