Tao Li, Parth Anand Jawale, Martha Palmer, Vivek Srikumar
TL;DR本文提出了一种结构调整框架,以在训练时通过软化约束来提高模型性能,利用神经网络的表达能力和具有结构化损失的监督学习组件,通过实验证明可以在语义角色标注任务中取得比 RoBERTa 等基线更好的结果,并在低资源情况下实现了持续改进。
Abstract
Recent neural network-driven semantic role labeling (SRL) systems have shown
impressive improvements in F1 scores. These improvements are due to expressive
input representations, which, at least at the surface, a