TAT-DQA: Towards Complex Document Understanding By Discrete Reasoning

TAT-DQA is a large-scale Document VQA dataset, which is constructed by extending the TAT-QA. It aims to stimulate progress of QA research over more complex and realistic visually-rich documents with rich tabular and textual content, especially those requiring numerical reasoning.

You can download our TAT-DQA dataset via TAT-DQA Dataset.

For more information, please refer to our TAT-DQA Website or read our ACM MM 2022 paper PDF.

Updates

${\color{red}Jan 2024}$: We release the ground truth for the TAT-DQA test set TAT-DQA Dataset, to facilitate future research on this task!

${\color{red}May 2023}$: Source Code released! You are welcome to use the Doc2SoarGraph repo to explore the TAT-DQA dataset and start your research!

Citation

Please kindly cite our work if you use our dataset or codes, thank you.

@inproceedings{zhu2022towards,
  title={Towards complex document understanding by discrete reasoning},
  author={Zhu, Fengbin and Lei, Wenqiang and Feng, Fuli and Wang, Chao and Zhang, Haozhou and Chua, Tat-Seng},
  booktitle={Proceedings of the 30th ACM International Conference on Multimedia},
  pages={4857--4866},
  year={2022}
}

@misc{zhu2023doc2soargraph,
      title={Doc2SoarGraph: Discrete Reasoning over Visually-Rich Table-Text Documents with Semantic-Oriented Hierarchical Graphs}, 
      author={Fengbin Zhu and Chao Wang and Fuli Feng and Zifeng Ren and Moxin Li and Tat-Seng Chua},
      year={2023},
      eprint={2305.01938},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

License

The TAT-DQA dataset is under the license of Creative Commons (CC BY) Attribution 4.0 International

Any Questions?

For any issues please create an issue here or kindly email us at: Fengbin Zhu [email protected], thank you.