We have presented a new dataset for question and answering models. Our dataset contains 27 different Urdu paragraphs which are taken from different available resources i.e Urdu Wikipedia, youtube and news articles etc. All selected paragraphs have an average of 3 to 7 questions along with their possible answers that range from 1 to 3. The data contains mostly Urdu words as well as some words from English language.
| Type | Count |
|---|---|
| Total # of Paragraph | 27 |
| Total # of Question | 499 |
| Total # of worlds in all paragraphs | 5553 |
| Total # of unique worlds in all paragraphs | 1631 |
| Total # of worlds in all Question | 1237 |
| Total # of unique worlds in all Question | 395 |
Structure of dataset:
Assuming variable data has the data imported into it as a dictionary:
To access first question-context set: data['1']
To access all questions of first datapoint: data['1']['question']
To access context of first datapoint:
data['1']['context']
Any questions or comments or if you are willing to contribute?
Let us know at ahsan.farooqui@ieee.org