English

End-to-End Synthetic Data Generation for Domain Adaptation of Question Answering Systems

Computation and Language 2020-10-14 v1

Abstract

We propose an end-to-end approach for synthetic QA data generation. Our model comprises a single transformer-based encoder-decoder network that is trained end-to-end to generate both answers and questions. In a nutshell, we feed a passage to the encoder and ask the decoder to generate a question and an answer token-by-token. The likelihood produced in the generation process is used as a filtering score, which avoids the need for a separate filtering model. Our generator is trained by fine-tuning a pretrained LM using maximum likelihood estimation. The experimental results indicate significant improvements in the domain adaptation of QA models outperforming current state-of-the-art methods.

Keywords

Cite

@article{arxiv.2010.06028,
  title  = {End-to-End Synthetic Data Generation for Domain Adaptation of Question Answering Systems},
  author = {Siamak Shakeri and Cicero Nogueira dos Santos and Henry Zhu and Patrick Ng and Feng Nan and Zhiguo Wang and Ramesh Nallapati and Bing Xiang},
  journal= {arXiv preprint arXiv:2010.06028},
  year   = {2020}
}

Comments

EMNLP 2020