Distant Supervised Relation Extraction Model for Reinforcement Learning Combined with Noise Network

Abstract

The distant supervised relation extraction has received wide attention from scholars in recent years. Existing methods for distant supervised relation extraction are based on bag-level for relation prediction, but they do not correspond to sentences and relation one by one. So in 2018, some scholars have proposed distant supervised relation extraction based on the sentence-level with reinforcement learning methods. Adding adaptive noise to the parameters of the reinforcement learning algorithm can effectively improve the performance of the algorithm. In this paper, the parametric noise is added to the neural network weights to increase the exploration of reinforcement learning. Experiments show that reinforcement learning with noise effectively improves the effect of distant supervised relation extraction based on the sentence-level.