Apply Bert-based models and Domain knowledge for Automated Legal Question Answering tasks at ALQAC 2021

Abstract

With robust development in NLP (Natural Language Processing) methods and Deep Learning, there are a variety of solutions to the problems in question answering systems that achieve extraordinary results. In this paper, we describe our approach using at the Automated Legal Question Answering Competition (ALQAC) 2021. In this competition, we achieved the first prize of all tasks with the scores of 88.07%, 71.02%, 69.89% in Task 1, Task 2 and Task 3 respectively.