The Research Of Chinese Relation Extraction Based On Deep Learning

Relation extraction is the detection and identification of semantic relations between natural language text entities.Relation extraction has high significance for many applications of NLP,such as question answering,knowledge graphs,etc.Recently,deep learning models have emerged as powerful tools for relation extraction.However,little work has been done on relation extraction for the Chinese language.One major challenge for relation extraction in Chinese texts is that Chinese sentences have no obvious word segmentation.This ambiguity increases the possibility of word segmentation errors.Another challenge is the lack of Chinese text datasets.Firstly,based on the idea of distant supervision,the relationship triples were extracted from Fudan Knowledge Factory and Baidu Encyclopedia.By aligning with the text of the open Chinese corpus Souhu news dataset,a Chinese character relations hip data set was constructed.In this paper,we propose an attention-based multi-instance multilabel bidirectional long short-term memory network for distantly supervised Chinese relation extraction.Our model takes Chinese character embeddings and positi on embeddings as input without Chinese word segmentation errors.Then,the attention mechanism is used to extract richer Chinese character and sentence features.Finally,we handle the multi-label nature of relation extraction by using multi-label loss functions in the neural network classifier.In order to improve the performance of the relation extraction model,an improved relational pooling BGRU-based relationship extraction model is proposed,which replaces LSTM with a simpler GRU of neuron structure and adds segmentation after the character-level attention layer.The piecewise pooling layer combines the features of RNN to capture global information and the maximum pooling of segments to extract important information,while the classifier uses the commonly used softmax classification function.Experiments on the constructed Chinese dataset show that the two neural network models proposed in this paper are suitable for Chinese relation extraction and have high performance.In addition,experiments were ca rried out on an English benchmark dataset.The results show that the proposed methods is superior to some existing methods.