A Robust Adversarial Training Approach to Machine Reading Comprehension

本文从阅读理解模型自身出发，利用对抗的思路自动发现对抗样本，从而有针对性地加强模型的鲁棒性。 paper: https://drive.google.com/open?id=1lWlaRQvSW8orUwmIM5HRRKtp_K2gQKessource: AAAI 2020

NLP

2020-03-02

Reading Turn by Turn：Hierarchical Attention Architecture for Spoken Dialogue Comprehension

本文研究的是对话阅读理解问题，提出了层次注意力结构（turn-level+word-level）来促进对话理解。实验证明了本文采用的turn-based的建模方式优于utterance-based，并且在长对话以及包含未知实体的对话中（论文称之为out-of-distribution）优势更加明显。 paper: https://drive.google.com/open?id=17nTPUASgBGx7g3GDRyHbyvqzRdlm4xa9source: ACL 2019 short paper

Helic He

Tag: MRC

A Robust Adversarial Training Approach to Machine Reading Comprehension

Reading Turn by Turn：Hierarchical Attention Architecture for Spoken Dialogue Comprehension