Talk to Paper ACL2020

Official repository for Talk to Paper demo at ACL2020 [https://www.aclweb.org/anthology/2020.acl-demos.5/]

References

If you use any source codes or datasets included in this toolkit in your work, please cite the following paper. The bibtex are listed below:

@article{zhao2020talk,
  title={Talk to Papers: Bringing Neural Question Answering to Academic Search},
  author={Zhao, Tianchang and Lee, Kyusong},
  journal={arXiv preprint arXiv:2004.02002},
  year={2020}
}

Try the system

You can try the demo at https://ask.soco.ai

Index your own data

You can sign-up for free task at https://app.soco.ai to index your own data. If your index size exceeds the limit, please contact us at [contact@soco.ai] for special offer to researchers and students.

Dataset

There are two types of data:

Question-only Data

the first type is question only data that is not linked to specific answer spans in the paper PDF. This question only dataset will be included at \data

Question-to-answer Data

Due to the challenges of collecting question-answer pairs for academia papers (it requires the annotator to have siginificant expertise in AI/NLP), we have made it possible for anyway to annotate answers and linked questions at https://ask.soco.ai.