AI Rule Generation from Natural Language Text via Structured Knowledge
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current knowledge acquisition methods for artificial intelligence systems rely heavily on human-defined rules or supervised black-box modeling, limiting the ability of AI to self-learn rules that describe natural language segments in terms of structured knowledge.
Innovation Solution
A system comprising a processor and memory that uses a masking component to associate unmasked elements of a natural language text segment with structured knowledge elements, a prediction component to predict masked elements based on extended context, and a scoring component to calculate estimated scores, facilitating self-learning of rules that describe natural language text segments in terms of structured knowledge.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If human-defined rules or supervised black-box modeling are used for knowledge acquisition, then the AI system can process natural language text, but the system cannot autonomously learn rules and requires continuous human intervention
Solution Approach 1:
The system enables autonomous rule learning by allowing the AI model to self-generate rules from natural language text without human intervention. The rule generator automatically creates structured knowledge representations, and the loss function autonomously evaluates and refines these rules through iterative training, eliminating the need for continuous human-defined rule updates
Solution Approach 2:
The patent introduces an intermediary rule generation mechanism that bridges natural language text and structured knowledge bases. This intermediary component translates unstructured text into formal logical representations, enabling automatic knowledge acquisition without requiring direct human intervention in the knowledge base construction process
2Reliability
If the knowledge base is expanded with more structured knowledge elements, then the AI system's factual correctness improves, but the complexity of maintaining and updating the knowledge base increases
Solution Approach 1:
The system automatically maintains and updates the knowledge base by continuously generating new rules from incoming natural language text. The autonomous rule generation process expands the knowledge base without requiring manual curation, and the built-in loss function automatically evaluates factual correctness through consistency checking against existing knowledge
Solution Approach 2:
The system performs preliminary validation and consistency checking during the rule generation process itself. By evaluating candidate rules against existing knowledge bases before full integration, the system prevents factual errors and maintains consistency proactively, reducing the need for later correction and maintenance efforts
3Adaptability or versatility
If supervised black-box modeling is used, then the AI system can learn patterns from data, but the system lacks interpretability and cannot generate explicit rules
Solution Approach 1:
The patent segments the learning process into distinct components: pattern recognition from natural language text, rule generation from identified patterns, and structured knowledge representation. This segmentation allows the system to maintain both the adaptability of pattern recognition and the interpretability of explicit rules, as each component can be optimized independently while contributing to the overall system
Solution Approach 2:
The rule generation component serves as an intermediary that translates learned patterns into explicit, interpretable rules. This intermediary layer preserves the pattern recognition capabilities while producing human-readable logical representations, preventing the loss of interpretability that occurs in pure black-box models
Data Source
AI summary
A system can comprise a memory that stores computer executable components, and a processor, operably coupled to the memory, that executes the computer executable components comprising: a linking component that associates one or more unmasked elements of the logical form with one or more corresponding structured knowledge elements of a knowledge base and a prediction component that predicts the one or more masked elements based on extended context of the corresponding structured knowledge elements of the knowledge base to generate one or more predicted elements. In an embodiment, the prediction component predicts the one or more masked elements based on scores of one or more candidate elements. In an embodiment, the system can determine one or more rules that describe the natural language text segment in terms of the structured knowledge elements and associated weights of the knowledge base paths.