Sentence compression with reinforcement learning

Abstract

Deletion-based sentence compression is frequently formulated as a constrained optimization problem and solved by integer linear programming (ILP). However, ILP methods searching the best compression given the space of all possible compressions would be intractable when dealing with overly long sentences and too many constraints. Moreover, the hard constraints of ILP would restrict the available solutions. This problem could be even more severe considering parsing errors. As an alternative solution, we formulate this task in a reinforcement learning framework, where hard constraints are used as rewards in a soft manner. The experiment results show that our method achieves competitive performance with a large improvement on the speed.