README

The training data is a json file with code-mixed questions and answers.

The languages present in the dataset include Hinglish, Tenglish and Tamlish.

The file includes the keys 'user', which you can change to your username while submitting the test batches. More information about this will be updated soon.

The 'division' key corresponds to 'train'/'test' divisions.

The questions are elicited from hinglishpedia.com and general images from Google Search.

A dictionary with key 'questions', has a list of questions. Each element in the list has the following information: 
'context': hinglishpedia article or general for image questions
'query': the body of the question
'answer': answer
'id': question id
and language information.


If you use this data for a publication or other work, please cite the following work: (A conference paper corresponding to this work is onging, please recheck the citation here):

@article{chandu2018code,
  title={Code-Mixed Question Answering Challenge: Crowd-sourcing Data and Techniques},
  author={Chandu, Khyathi Raghavi and Loginova, Ekaterina and Gupta, Vishal and van Genabith, Josef and {\"u}nter Neuman, G and Chinnakotla, Manoj and Nyberg, Eric and Black, Alan},
  journal={ACL 2018},
  pages={29},
  year={2018}
}