BriefGPT.xyz
Jul, 2017
神经误差检测模型的辅助目标
Auxiliary Objectives for Neural Error Detection Models
HTML
PDF
Marek Rei, Helen Yannakoudakis
TL;DR
通过神经序列标记方法探究不同的辅助目标和训练策略在学习者写作错误检测中的实用性；实验结果表明，采用多目标联合学习和领域内平行标签可以提高模型性能，而不增加参数量。
Abstract
We investigate the utility of different
auxiliary objectives
and
training strategies
within a
neural sequence labeling
approach to
→