A Minimal Approach for Natural Language Action Space in Text-based Games
File version
Version of Record (VoR)
Author(s)
Fang, M
Haffari, G
Pan, S
Shareghi, E
Griffith University Author(s)
Primary Supervisor
Other Supervisors
Editor(s)
Date
Size
File type(s)
Location
Singapore, Singapore
Abstract
Text-based games (TGs) are language-based interactive environments for reinforcement learning. While language models (LMs) and knowledge graphs (KGs) are commonly used for handling large action space in TGs, it is unclear whether these techniques are necessary or overused. In this paper, we revisit the challenge of exploring the action space in TGs and propose 𝜖-admissible exploration, a minimal approach of utilizing admissible actions, for training phase. Additionally, we present a text-based actor-critic (TAC) agent that produces textual commands for game, solely from game observations, without requiring any KG or LM. Our method, on average across 10 games from Jericho, outperforms strong baselines and state-of-the-art agents that use LM and KG. Our approach highlights that a much lighter model design, with a fresh perspective on utilizing the information within the environments, suffices for an effective exploration of exponentially large action spaces.
Journal Title
Conference Title
Proceedings of the 27th Conference on Computational Natural Language Learning (CoNLL)
Book Title
Edition
Volume
Issue
Thesis Type
Degree Program
School
Publisher link
Patent number
Funder(s)
Grant identifier(s)
Rights Statement
Rights Statement
© 2023 Association for Computational Linguistics. The ACL materials that are hosted in the Anthology are licensed to the general public under a liberal usage policy that allows unlimited reproduction, distribution and hosting of materials on any other website or medium, for non-commercial purposes. Prior to 2016, all ACL materials are licensed using the Creative Commons 3.0 BY-NC-SA (Attribution, Non-Commercial, Share-Alike) license. As of 2016, this policy has been relaxed further, and all subsequent materials are available to the general public on the terms of the Creative Commons 4.0 BY (Attribution) license; this means both commercial and non-commercial use is explicitly licensed to all.
Item Access Status
Note
Copyright permissions for this publication were identified from the publisher's website at https://aclanthology.org/faq/
Access the data
Related item(s)
Subject
Artificial intelligence
Computer gaming and animation
Persistent link to this record
Citation
Ryu, DK; Fang, M; Haffari, G; Pan, S; Shareghi, E, A Minimal Approach for Natural Language Action Space in Text-based Games, Proceedings of the 27th Conference on Computational Natural Language Learning (CoNLL), 2023, pp. 138-154