Generative AI Safeguarding via Ontological Concept Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing generative AI models lack the ability to validate the accuracy of translated, simplified, or summarized content, particularly in domains like healthcare, where important information may be omitted or fabricated.
Innovation Solution
A process that applies safeguards to generative AI tasks by using multi-lingual information extraction and task-specific rules based on standard ontologies and taxonomies to detect hallucinations and omissions in the target content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If generative AI models are used to translate, simplify, or summarize clinical texts, then productivity is improved, but reliability deteriorates due to inability to validate accuracy, detect omissions, or identify hallucinations
Solution Approach 1:
The patent introduces an intermediary validation system that sits between the generative AI model and the final output. This system extracts ontological concepts from both source and target content, compares them using a comparison score, and identifies discrepancies such as omissions and hallucinations. The intermediary validator enables automated accuracy checking without requiring manual review, thus maintaining productivity while improving reliability.
2Ease of operation
If existing approaches are used for content translation and summarization, then ease of operation is improved, but measurement precision deteriorates due to lack of validation capabilities
Solution Approach 1:
The validation system performs self-service by automatically extracting ontological concepts, comparing source and target content, and identifying issues without requiring external manual verification. The system uses pre-defined ontological frameworks and comparison scores to autonomously validate accuracy, maintaining ease of operation while achieving precise measurement of output quality.
3Device complexity
If generative AI outputs are used without validation, then device complexity is reduced, but loss of information increases due to omissions and hallucinations
Solution Approach 1:
The system performs preliminary extraction of ontological concepts from source content before the generative AI processing occurs. By having these reference concepts ready in advance, the validation process can efficiently compare them against the target content afterward, detecting omissions and hallucinations. This preliminary preparation enables comprehensive information validation without significantly increasing overall system complexity.
Data Source
AI summary
A method, computer program product, and computing system for processing target content generated by processing source content using a generative artificial intelligence (AI) model, where the generative AI model performs a task using the source content to generate the target content. An ontological concept is extracted from the source content using a natural language processing (NLP) engine. An ontological concept is extracted from the target content using the NLP engine. An ontological concept comparison score is generated by comparing the ontological concept from the source content and the ontological concept from the target content based upon, at least in part, the task performed using the source content to generate the target content. An issue is identified in the target content based upon, the ontological concept comparison score and the task performed using the source content to generate the target content.