An Empirical Study of Vietnamese Machine Reading Comprehension with Unsupervised Context Selector and Adversarial Learning

EasyChair Preprint 7073

Abstract

Machine Reading Comprehension (MRC) is a great NLP task that requires focus on making the machine read, scan documents, and extract meaning from the text, just like a human reader. One of the MRC system challenges is not only having to understand the context to extract the answer but also being aware of the trust-worthy of the given question is possible or not. Pretrained language models (PTMs) that are based on transformer architecture has proved their performance on lots of NLP downstream tasks but it still has limitation in the fixed-length input. We propose an unsupervised context selector that shortens given the context but still contains the answers within related contexts. In VLSP2021-MRC shared task (Nguyen et al., 2021), we also empirical several training strategies including unanswerable question sample selection and different adversarial training approaches.

Keyphrases: Machine Reading Comprehension, VLSP2021-MRC Shared Task, adversarial training

@booklet{EasyChair:7073,
  author    = {Hoang Vu Tran and Phuc Minh Nguyen},
  title     = {An Empirical Study of Vietnamese Machine Reading Comprehension with Unsupervised Context Selector and Adversarial Learning},
  howpublished = {EasyChair Preprint 7073},
  year      = {EasyChair, 2021}}