Joint Entity and Relation Extraction with Adaptive Thresholding.

Entity naming recognition and relation extraction are two very important subtasks in natural language processing tasks, and traditional processes do the two tasks separately, which may ignore the dependencies and associations between the two tasks. With the development of deep learning, joint entity and relation extraction based on deep learning has received wide scholarly attention, but the joint extraction task still faces challenges such as unclear entity boundaries and overlapping entities or relations. In this paper, we innovatively design a joint extraction model based on a multi-headed self-attentive mechanism, and also introduce a word-level adaptive threshold to replace the global threshold to help the model extract triples. To verify the effectiveness of our model, we conduct experiments on two publicly available datasets, and the experimental results show that our model has good performance.