Hybrid AI System Bridges Natural Language and Answer Set Programming Using LLMs and AMR

A new hybrid AI system aims to bridge the gap between natural language and Answer Set Programming (ASP), a powerful declarative programming paradigm for solving combinatorial problems. Developed by researchers Connar Hite, Sean Saud, Raef Taha, Nayim Rahman, Tanvir Atahary, Scott Douglass, and Tarek Taha, the system uniquely combines Large Language Models (LLMs) and Abstract Meaning Representation (AMR) to translate unconstrained English into ASP programs. The system minimizes the role of LLMs to tasks like simplifying sentences, identifying keywords, and generating simple facts. AMR graphs are then parsed from the simplified language to systematically generate ASP constraints. This approach seeks to mitigate the limitations of purely LLM-based systems, such as hallucination, bias, and lack of explainability. The system has been successfully demonstrated on example logic puzzles, generating complete, runnable ASP programs that represent and solve the original problem. This hybrid approach represents a significant step towards creating a lighter-weight, more explainable system that allows users unfamiliar with programming to interact with code and solve complex logic problems using natural language.