t5-base-table-question-generator

Model description

This is an t5-base model, finetuned to generate questions given a table using WikiSQL dataset. It was trained to take the SQL, answer and column header of a table as input to generate questions. For more information check our T3QA paper from EMNLP 2021.

Overview

Language model: t5-base
Language: English
Task: Table Question Generation
Data: WikiSQL

Intented use and limitations

One can use this model to generate questions given a table. Biases associated with pre-training of T5 and WikiSQL dataset may be present.

Usage

One can use this model directly in the PrimeQA framework as in this example notebook.

Citation

@inproceedings{chemmengath2021topic,
  title={Topic Transferable Table Question Answering},
  author={Chemmengath, Saneem and Kumar, Vishwajeet and
          Bharadwaj, Samarth and Sen, Jaydeep and
          Canim, Mustafa and Chakrabarti, Soumen and
          Gliozzo, Alfio and Sankaranarayanan, Karthik},
  booktitle={Proceedings of the 2021 Conference on 
      Empirical Methods in Natural Language Processing},
  pages={4159--4172},
  year={2021}
}