Named Entity Recognition Method of Brazilian Legal Text based on pre-training model
Zhili Wang, Yufan Wu, Pengbin Lei, Cheng Peng · Journal of Physics Conference Series · 2020
Abstract Named entity recognition (NER) is a common task in Natural Language Processing (NLP). To this end, we propose a novel approach based on pre-training model to complete the sequence labeling tasks by learning the large-scale real-world data from Brazilian legal documents. Especially, combining iterated dilated convolution[1] (IDCNN) and Bi-LSTM, we develop the scalable sequence labeling model named Sequence Tagging Model (STM) and extensive experiments validate the effectiveness of STM for NER tasks. Furthermore, compared with the IDCNN-CRF model, the experimental results show that the STM is better and the F1 score is 93.23%, which provides an important basis for NER tasks.