LexGLUE

与 GLUE 和 SuperGLUE （ Wang et al.， 2109） 一样，我们的目标之一是推动能够处理多个 NLP 任务的通用（或基础）模型，在我们的例子中是合法的 NLP 任务，可能具有有限的特定任务微调。另一个目标是为希望探索或开发法律NLP方法的NLP研究人员和从业者提供一个方便且信息丰富的切入点。考虑到这些目标，我们在 LexGLUE 中包含的数据集及其处理的任务已通过多种方式进行了简化，如下所述，以使新手和通用模型更容易解决所有任务。One of our goals, similar to GLUE and SuperGLUE (Wang et al., 2109), is to advance general-purpose (or foundational) models capable of handling multiple NLP tasks, in our case legal NLP tasks, potentially with limited task-specific fine-tuning. Another goal is to provide a convenient and informative entry point for NLP researchers and practitioners who wish to explore or develop legal NLP methods. With these goals in mind, the datasets included in LexGLUE and the tasks they address have been simplified in multiple ways, as described below, to make it easier for novices and general-purpose models to solve all tasks.

搜集汇总

背景与挑战

背景概述

LexGLUE是一个用于英语法律语言理解的基准数据集，基于七个现有的法律NLP数据集，涵盖多标签分类、多类分类和多项选择QA等任务，旨在推动通用模型在法律领域的应用，并为研究人员提供便捷的实验入口。