An Internet Resource Extracting Relation Model based on Distant Supervision and Deep Learning

Relational extraction of large-scale text data is one of the key issues in natural language processing. Facing large-scale Internet data resources, this paper proposes a relationship extraction model based on distant supervision and deep learning. The model build entity-to-package based on distant supervision to solve the high computational complexity and time cost of Internet large-scale data. Model's input embeddings based on BERT and entity position, which can improve the text feature representation ability. And the model utilized attention mechanism to optimize the entity-to-package tag attention parameters to alleviate the wrong label problems arising from distant supervision learning. In experiments, the performance of the proposed model exceeds that of other baseline models. The results demonstrate the positive contribution of our model in Internet large-scale relation extraction.