UniK-QA: Unified Representations of Structured and Unstructured Knowledge for Open-Domain Question Answering

This repo contains code for our paper (UniK-QA is pronounced as unique-QA):

UniK-QA: Unified Representations of Structured and Unstructured Knowledge for Open-Domain Question Answering
Barlas Oguz*, Xilun Chen*, Vladimir Karpukhin, Stan Peshterliev, Dmytro Okhonko, Michael Schlichtkrull, Sonal Gupta, Yashar Mehdad, Scott Yih
(*Equal Contribution)
Meta AI

Getting the Data

After cloning the repo, run the following in the root directory of the repo to fetch the WebQSP data:

wget https://dl.fbaipublicfiles.com/UniK-QA/data.tar.xz
tar -xvf data.tar.xz

Reproducing the WebQSP experiment

Follow the interactive Jupyter notebook UniK-QA on WebQSP.ipynb to reproduce our experiment on WebQSP.

• The script will first convert the KB relations into text sentences.
• DPR is then run to select the most relevant relations for each question.
• Next, the input to the FiD reader is created for each question using the most relevant relations retrieved by DPR.
• Finally, a FiD model can be trained using the UniK-QA input. Our trained FiD checkpoint can be downloaded here. (Our model was trained in late 2020, so you may need to check out an older version of FiD.)

License

UniK-QA is CC-BY-NC 4.0 licensed, as found in the LICENSE file.