The study analyzes the relationship extraction in information extraction,and proposes an information extraction model based on deep learning to solve the problem that it is difficult to capture deeper semantic information between texts in existing relationship extraction technologies.Then the study conducts word vector embedding with BERT model and dependency graph through graph convolutional network,assigns different weights to different features with attention mechanism,completes the relationship classification by Softmax,realizes the relationship extraction,and conducts experimental comparison with 5 benchmark models in DocRED dataset.F1 value of the proposed model is proved to be the best,indicating that the model can improve the accuracy of extraction.
维普
万方数据