Contract Segment Processing Using Embedding Similarity

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

The process of preparing, reviewing, and negotiating contracts is time-consuming and prone to errors, especially when dealing with contract segments from different templates or languages, which increases the risk of misinterpretation and non-compliance.

Innovation Solution

A computer-implemented method for processing contract documents by receiving a segment from a contract document, accessing approved segments to find the most similar one using a similarity metric, and using that approved segment for further processing, such as automatic amendment proposals and risk scoring.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Extent of automation

If template clause language and red lining approaches are used, then contract processing can be automated to some extent, but the system becomes difficult to implement when dealing with contracts from different parties using different templates or languages

Engineering Contradiction:

Improveautomation of contract processingVSAvoidadaptability to different templates and languages

Core Design Contradiction:

Extent of automationVSAdaptability or versatility

Solution Approach 1:

The system transforms contract segments into numerical embeddings (vector representations) that capture semantic meaning. This parameter transformation allows the system to compare contracts regardless of their original template or language, enabling automated processing while maintaining adaptability to different contract formats and languages through mathematical representation rather than rigid template matching

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent introduces an intermediary embedding model that translates different contract templates and languages into a common numerical representation space. This intermediary layer enables comparison and processing of diverse contracts without requiring the system to understand each specific template or language, thus maintaining both automation and versatility

Inventive Principle:

Principle #24Intermediary (Mediator)

2Productivity

If generative AI approaches are used, then processing time is reduced, but the system may depart from template language requiring additional manual processing

Engineering Contradiction:

Improveprocessing speedVSAvoidcompliance with template language requirements

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The system employs feedback mechanisms where generated contract segments are compared against approved segments using similarity metrics. This feedback loop ensures that generative AI outputs remain compliant with organizational template requirements by continuously referencing and aligning with approved language patterns, thus maintaining reliability while benefiting from AI-driven speed

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system performs preliminary actions by pre-processing and storing approved contract segments as reference embeddings before generation occurs. This preliminary preparation enables the generative AI to quickly generate compliant segments by referencing pre-processed approved language, maintaining both speed and template compliance without requiring post-generation manual review

Inventive Principle:

Principle #10Preliminary action

3Reliability

If manual review and negotiation of each contract segment is performed, then accuracy and compliance are maintained, but the process becomes time-consuming and error-prone

Engineering Contradiction:

Improveaccuracy and complianceVSAvoidtime for review and negotiation

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system enables self-service by automatically comparing received contract segments against approved segments using similarity metrics and embedding comparisons. This automated self-assessment maintains accuracy and compliance by referencing approved language while eliminating time-consuming manual review, as the system independently determines whether segments require human attention

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system applies partial action by performing automated compliance checking on all segments but requiring full manual review only for segments that fall below a similarity threshold. This approach maintains reliability for critical segments while reducing time loss for segments that are already compliant, avoiding the need for exhaustive manual review of every contract segment

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20250068850A1Document processing

Publication Date: 2025.02.27 IVALUA S AS

AI summary

There is provided a computer implemented method of processing contract documents. The method comprises receiving a segment associated with a received contract document and for the received segment, accessing one or more approved segments to determine an approved segment from the one or more approved segments that is similar to the received segment, wherein the approved segment is determined using a similarity metric between the received segment and at least some of the approved segments. Further processing of the received segment using the determined approved segment.