MQR

The is the repository for the paper How to Ask Better Questions? A Large-Scale Multi-Domain Dataset for Rewriting Ill-Formed Questions

Train Data

• The training data can be downloaded from: https://drive.google.com/drive/folders/1VZ7BusNjbulG9xYlcbGmip5Pl-Gzo8HD?usp=sharing

Dev/Test Data and Model Predictions

• The dev and test datasets are under directory data
• The rewritten dev/test splits can be found under each subdirectory of data

Annotation

• All annotations on the DEV/TEST splits described in the paper can be found at https://docs.google.com/spreadsheets/d/1uPDhGzxia2T0t9WXjMDuopAtgk6CceZ1U3mrDFNI4cY/edit?usp=sharing

License

The MQR dataset is under cc-by-sa 4.0 license, intended to be shared and remixed.

• https://creativecommons.org/licenses/by-sa/4.0/

The MQR dataset is partially constructed from the Stack Exchange data dumps

• https://archive.org/details/stackexchange

We used Quora Question Pairs dataset as part of the training data

• https://www.quora.com/q/quoradata/First-Quora-Dataset-Release-Question-Pairs
• https://www.quora.com/about/tos

We also used the Paralex dataset for training

• http://knowitall.cs.washington.edu/paralex/

Reference

@inproceedings{chu-mqr-20,
  author    = {Zewei Chu and Mingda Chen and Jing Chen and Miaosen Wang and Kevin Gimpel and Manaal Faruqui and Xiance Si},
  title     = {How to Ask Better Questions? A Large-Scale Multi-Domain Dataset for Rewriting Ill-Formed Questions},
  booktitle = {Proc. of {AAAI}},
  year      = {2020}
}