English

Stronger Transformers for Neural Multi-Hop Question Generation

Computation and Language 2020-10-23 v1 Machine Learning

Abstract

Prior work on automated question generation has almost exclusively focused on generating simple questions whose answers can be extracted from a single document. However, there is an increasing interest in developing systems that are capable of more complex multi-hop question generation, where answering the questions requires reasoning over multiple documents. In this work, we introduce a series of strong transformer models for multi-hop question generation, including a graph-augmented transformer that leverages relations between entities in the text. While prior work has emphasized the importance of graph-based models, we show that we can substantially outperform the state-of-the-art by 5 BLEU points using a standard transformer architecture. We further demonstrate that graph-based augmentations can provide complimentary improvements on top of this foundation. Interestingly, we find that several important factors--such as the inclusion of an auxiliary contrastive objective and data filtering could have larger impacts on performance. We hope that our stronger baselines and analysis provide a constructive foundation for future work in this area.

Keywords

Cite

@article{arxiv.2010.11374,
  title  = {Stronger Transformers for Neural Multi-Hop Question Generation},
  author = {Devendra Singh Sachan and Lingfei Wu and Mrinmaya Sachan and William Hamilton},
  journal= {arXiv preprint arXiv:2010.11374},
  year   = {2020}
}

Comments

Code will be made available

R2 v1 2026-06-23T19:32:21.819Z