基于Bert的中文法律术语角色识别

Role Recognition of Legal Terms Based on BERT

针对法律术语角色识别问题,笔者提出了一种基于BERT预训练语言模型的方法进行法律术语角色识别,并制定了标注规范。由于BERT预训练语言模型本身含有丰富的语义特征,可以一定程度上解决训练语料不足的问题,所以首先通过BERT预训练模型将字向量送入下游任务中,再通过BiLSTM-CRF进行法律术语角色识别。实验结果表明,笔者提出的方法要优于对照组模型,具有重要的现实意义。

Aiming at the problem of role recognition of legal terms,the author proposes a method based on the Bert pre training language model to identify the legal terms role,and formulates the annotation specification.Because of the rich semantic features of the pre training language model itself,it can solve the problem of insufficient training corpus to a certain extent.Therefore,firstly,the word vector is sent into the downstream task through the Bert pre training model,and then the legal term role is recognized through bilstm-crf.The experimental results show that the method proposed by the author is superior to the control group model and has important practical significance.