Dataset Card for Spivavtor

Paper: Spivavtor: An Instruction Tuned Ukrainian Text Editing Model

Authors: Aman Saini, Artem Chernodub, Vipul Raheja, Vivek Kulkarni

Dataset Summary

This is the dataset used to train all Spivavtor models. It contains data for 4 tasks - Grammatical Error Correction (GEC), Simplification, Coherence and Paraphrasing. The specific details are as follows:

Task	Examples in Training data	Examples in Validation data
GEC	27,929	3,103
Simplification	11,501	1,278
Coherence	9,278	1,031
Paraphrasing	14,076	1,564
Total	62,784	6,976

Full details of the dataset can be found in our paper.

Dataset Structure

The dataset is in JSON format.

Data Instances

{
  "id": 1,
  "src": "Перефразуйте: Кайс став одержимий нещодавно усвідомленою вразливістю.",
  "tgt": "Кайс був одержимий своєю нововиявленою вразливістю.",
  "task": "paraphrase"
}

Data Fields

• id: Unique ID for an instance
• src: Input text in Ukrainian (formatted as instruction: input_text)
• tgt: Output text in Ukrainian
• task: Text editing task for this instance