Stochastic Policy

Supervised approaches for text summari- sation suffer from the problem of mis- match between the target labels/scores of individual sentences and the evaluation score of the final summary. Reinforcement learning …...