“…Over the past several years, the deep learning-based entity relationship extraction has become a research hotspot, as it alleviates the problem of error labelling and feature extraction error propagation in a distant supervised data set. The approaches mainly include convolutional neural networks (Li et al , 2020a, 2020b; Santos et al , 2015; Zeng et al , 2014), recurrent neural networks (Katiyar and Cardie, 2017; Tourille et al , 2017), generative adversarial networks (Qin et al , 2018), deep reinforcement learning (Xiao et al , 2020) and so on. For example, Zeng et al (2014) proposed using the word vector and the word position vector as the input of the convolutional neural network (CNN) and introduced the distance information between the entity and other words, which can well take the entity information in the sentence into account for relation extraction.…”