IWPT 2009 - Dependency Parsing with Energy-based Reinforcement Learning

Dependency Parsing with Energy-based Reinforcement Learning

lidan zhang

11th International Conference on Parsing Technology (IWPT 2009)
Paris, France, 7th-9th October, 2009

Summary

We present a model which integrates dependency parsing with reinforcement learning based on Markov decision process. At each time step, a transition is picked up to construct the dependency tree in terms of the long-run reward. The optimal policy for choosing transitions can be found with SARSA algorithm. In SARSA, an approximation of the state-action function can be obtained by calculating the negative free energies for the Resticted Boltzmann Machine. The experimental results on CoNLL-X multilingual data show that the proposed model achieves comparable results with the current state-of-the-art methods.

START Conference Manager (V2.56.8 - Rev. 780)