LOA: Logical Optimal Actions for Text-based Interaction Games

TL;DR

LOA introduces a neuro-symbolic reinforcement learning framework for text-based interaction games, combining neural networks with symbolic knowledge to improve interpretability and performance, supported by an open-source Python implementation.

Contribution

This work presents the LOA framework, integrating neural and symbolic methods for decision-making in text-based games, with a focus on interpretability and comparative evaluation.

Findings

LOA outperforms some existing neuro-symbolic approaches.

LOA enhances interpretability of decision rules.

Open-source code facilitates further research.

Abstract

We present Logical Optimal Actions (LOA), an action decision architecture of reinforcement learning applications with a neuro-symbolic framework which is a combination of neural network and symbolic knowledge acquisition approach for natural language interaction games. The demonstration for LOA experiments consists of a web-based interactive platform for text-based games and visualization for acquired knowledge for improving interpretability for trained rules. This demonstration also provides a comparison module with other neuro-symbolic approaches as well as non-symbolic state-of-the-art agent models on the same text-based games. Our LOA also provides open-sourced implementation in Python for the reinforcement learning environment to facilitate an experiment for studying neuro-symbolic agents. Code: https://github.com/ibm/loa