Language Understanding for Text-based Games Using Deep Reinforcement\n Learning

2015/06/30 by Karthik Narasimhan, Tejas Kulkarni, Narasimhan, Karthik +3 · 2 voices · 22 citations
Computer Science · #Topic Modeling

paper · pdf · doi:10.48550/arxiv.1506.08941

Abstract

In this paper, we consider the task of learning control policies for\ntext-based games. In these games, all interactions in the virtual world are\nthrough text and the underlying state is not observed. The resulting language\nbarrier makes such environments challenging for automatic game players. We\nemploy a deep reinforcement learning framework to jointly learn state\nrepresentations and action policies using game rewards as feedback. This\nframework enables us to map text descriptions into vector representations that\ncapture the semantics of the game states. We evaluate our approach on two game\nworlds, comparing against baselines using bag-of-words and bag-of-bigrams for\nstate representations. Our algorithm outperforms the baselines on both worlds\ndemonstrating the importance of learning expressive representations.\n

Cited by

• TALES: Text Adventure Learning Environment Suite
• Agentic Entropy-Balanced Policy Optimization
• Deep Reinforcement Learning for Programming Language Correction
• Generating Text with Deep Reinforcement Learning
• Learning Transferable Visual Models From Natural Language Supervision
• Language as an Abstraction for Hierarchical Deep Reinforcement Learning
• Deep Reinforcement Learning with a Natural Language Action Space
• STEP: Success-Rate-Aware Trajectory-Efficient Policy Optimization
• Grounded Language Learning in a Simulated 3D World
• Adaptive early classification of temporal sequences using deep reinforcement learning
• Inductive Bias and Language Expressivity in Emergent Communication
• Dialog-based Language Learning
• Hierarchical Deep Reinforcement Learning: Integrating Temporal Abstraction and Intrinsic Motivation
• e-QRAQ: A Multi-turn Reasoning Dataset and Simulator with Explanations
• ByteSized32Refactored: Towards an Extensible Interactive Text Games Corpus for LLM World Modeling and Evaluation
• End-to-End Goal-Driven Web Navigation
• Deep Recurrent Q-Learning for Partially Observable MDPs
• TextWorld: A Learning Environment for Text-based Games
• Learning Dynamic Belief Graphs to Generalize on Text-Based Games
• Reinforcement Learning for Machine Learning Engineering Agents
• Counting to Explore and Generalize in Text-based Games
• Interactive Recommender System via Knowledge Graph-enhanced Reinforcement Learning

Discussions

• Language Understanding for Text-based Games Using Deep Reinforcement Learning [pdf] [hn, 31 points, 6 comments]
• MIT AI bot learns aspects of language by autonomously playing text-games [hn, 2 points, 0 comments]

Related

• Language Expansion In Text-Based Games
• Using reinforcement learning to learn how to play text-based games
• Reading and Acting while Blindfolded: The Need for Semantics in Text Game Agents
• Keep CALM and Explore: Language Models for Action Generation in Text-based Games
• Algorithmic Improvements for Deep Reinforcement Learning applied to Interactive Fiction