Learning High-Level Planning from Text
作者： S.R.K. Branavan , Regina Barzilay , Tao Lei , Nate Kushman
DOI:
关键词:
摘要: Comprehending action preconditions and effects is an essential step in modeling the dynamics of world. In this paper, we express semantics precondition relations extracted from text terms planning operations. The challenge connection to ground language at level relations. This type grounding enables us create high-level plans based on abstractions. Our model jointly learns predict perform guided by those We implement idea reinforcement learning framework using feedback automatically obtained plan execution attempts. When applied a complex virtual world describing that world, our relation extraction technique performs par with supervised baseline, yielding F-measure 66% compared baseline's 65%. Additionally, show planner utilizing these significantly outperforms strong, unaware baseline -- successfully completing 80% tasks as 69% for baseline.
aclweb.org参考文章(5)
Malik Ghallab, Dana Nau, Paolo Traverso, Automated Planning: Theory & Practice ,(2004)
Damien Ernst, Arthur Louette, Introduction to Reinforcement Learning MIT Press. ,(1998)
Ronald J. Williams, Simple Statistical Gradient-Following Algorithms for Connectionist Reinforcement Learning Machine Learning. ,vol. 8, pp. 229- 256 ,(1992) , 10.1007/BF00992696
A.G. Barto, R.S. Sutton, Reinforcement Learning: An Introduction ,(1988)
Yishay Mansour, Satinder P. Singh, Richard S Sutton, David A. McAllester, Policy Gradient Methods for Reinforcement Learning with Function Approximation neural information processing systems. ,vol. 12, pp. 1057- 1063 ,(1999)