Language Understanding for Text-based Games Using Deep Reinforcement Learning
arXiv:1506.08941
Abstract
In this paper, we consider the task of learning control policies for text-based games. In these games, all interactions in the virtual world are through text and the underlying state is not observed. The resulting language barrier makes such environments challenging for automatic game players. We employ a deep reinforcement learning framework to jointly learn state representations and action policies using game rewards as feedback. This framework enables us to map text descriptions into vector representations that capture the semantics of the game states. We evaluate our approach on two game worlds, comparing against baselines using bag-of-words and bag-of-bigrams for state representations. Our algorithm outperforms the baselines on both worlds demonstrating the importance of learning expressive representations.
11 pages, Appearing at EMNLP, 2015
References in corpus (2)
Cited by in corpus (57)
- Learning Transferable Visual Models From Natural Language Supervision
- Learning to Communicate with Deep Multi-Agent Reinforcement Learning
- Hierarchical Deep Reinforcement Learning: Integrating Temporal Abstraction and Intrinsic Motivation
- Deep Recurrent Q-Learning for Partially Observable MDPs
- Deep Reinforcement Learning: An Overview
- Semi-supervised Deep Reinforcement Learning in Support of IoT and Smart City Services
- Deep Reinforcement Learning for Dialogue Generation
- Sample Efficient Actor-Critic with Experience Replay
- Survey on reinforcement learning for language processing
- Grounded Language Learning in a Simulated 3D World
- Towards End-to-End Learning for Dialog State Tracking and Management using Deep Reinforcement Learning
- Learn What Not to Learn: Action Elimination with Deep Reinforcement Learning
- TextWorld: A Learning Environment for Text-based Games
- Dialog-based Language Learning
- Evaluating Prerequisite Qualities for Learning End-to-End Dialog Systems
- Hierarchical and Interpretable Skill Acquisition in Multi-task Reinforcement Learning
- Conversational Contextual Cues: The Case of Personalization and History for Response Ranking
- Learning Dynamic Belief Graphs to Generalize on Text-Based Games
- Counting to Explore and Generalize in Text-based Games
- Model-Free Control for Distributed Stream Data Processing using Deep Reinforcement Learning
- Alignment of Language Agents
- Generating Text with Deep Reinforcement Learning
- Autonomous Quadrotor Landing using Deep Reinforcement Learning
- From Language to Programs: Bridging Reinforcement Learning and Maximum Marginal Likelihood
- Transforming Cooling Optimization for Green Data Center via Deep Reinforcement Learning
- Improving Information Extraction by Acquiring External Evidence with Reinforcement Learning
- Deep Reinforcement Learning with a Natural Language Action Space
- NAIL: A General Interactive Fiction Agent
- A Survey of Deep Learning Techniques for Mobile Robot Applications
- Deep Reinforcement Learning for Programming Language Correction
- Enhancing Text-based Reinforcement Learning Agents with Commonsense Knowledge
- LADDER: A Human-Level Bidding Agent for Large-Scale Real-Time Online Auctions
- Action Assembly: Sparse Imitation Learning for Text Based Games with Combinatorial Action Spaces
- Interactive Language Learning by Question Answering
- I love your chain mail! Making knights smile in a fantasy game world: Open-domain goal-oriented dialogue agents
- Reinforcement Learning with External Knowledge and Two-Stage Q-functions for Predicting Popular Reddit Threads
- Hierarchical Question Answering for Long Documents
- Language Expansion In Text-Based Games
- NLPGym -- A toolkit for evaluating RL agents on Natural Language Processing Tasks
- Reinforcement Learning based Multi-Access Control and Battery Prediction with Energy Harvesting in IoT Systems
- Jointly-Learned State-Action Embedding for Efficient Reinforcement Learning
- Probing Emergent Semantics in Predictive Agents via Question Answering
- Risk-Sensitive Deep RL: Variance-Constrained Actor-Critic Provably Finds Globally Optimal Policy
- Grounding Language for Transfer in Deep Reinforcement Learning
- Inductive Bias and Language Expressivity in Emergent Communication
- Interactive Recommender System via Knowledge Graph-enhanced Reinforcement Learning
- Towards Socially Intelligent Agents with Mental State Transition and Human Utility
- Mastering the Dungeon: Grounded Language Learning by Mechanical Turker Descent
- Inductive-bias-driven Reinforcement Learning For Efficient Schedules in Heterogeneous Clusters
- The Text-Based Adventure AI Competition
- Interactive Machine Comprehension with Information Seeking Agents
- Learning Representations and Agents for Information Retrieval
- A Coarse to Fine Question Answering System based on Reinforcement Learning
- Learning to Learn Semantic Parsers from Natural Language Supervision
- e-QRAQ: A Multi-turn Reasoning Dataset and Simulator with Explanations
- Playing optical tweezers with deep reinforcement learning: in virtual, physical and augmented environments
- Reinforcement Learning for Robust Missile Autopilot Design