/MemSum-DQA

Adapting an Efficient Long Document Extractive Summarizer for Document Question Answering

Primary LanguageJupyter NotebookApache License 2.0Apache-2.0

DOI

MemSum-DQA: Adapting an Efficient Long Document Extractive Summarizer for Document Question Answering

The code and data for the Kaggle Competition PDF-VQA of The first Workshop on Document Intelligence and Understanding (DocIU) in conjunction with the 32nd ACM International Conference on Information and Knowledge Management (CIKM 2023) in Birmingham, UK.

MemSum-DQA was ranked 1st in the Kaggle Competition PDF-VQA!

More details are available at https://www.kaggle.com/competitions/pdfvqa/overview and https://doc-iu.github.io/#three

End-to-end jupyter notebook: train_test_memsum_dqa.ipynb

Memsum-DQA