process/tweet_qg.py

import os
import json
from datasets import load_dataset

os.makedirs("data/tweet_qg", exist_ok=True)
data = load_dataset("lmqg/qg_tweetqa")


def process(tmp):
    tmp = [i.to_dict() for _, i in tmp.iterrows()]
    for i in tmp:
        i.pop('paragraph_question')
        i['text'] = f"context: {i.pop('paragraph')}, answer:{i.pop('answer')}"
        i['gold_label_str'] = i.pop('question')
    return tmp


train = process(data["train"].to_pandas())
val = process(data["validation"].to_pandas())
test = process(data["test"].to_pandas())
with open("data/tweet_qg/train.jsonl", "w") as f:
    f.write("\n".join([json.dumps(i) for i in train]))
with open("data/tweet_qg/validation.jsonl", "w") as f:
    f.write("\n".join([json.dumps(i) for i in val]))
with open("data/tweet_qg/test.jsonl", "w") as f:
    f.write("\n".join([json.dumps(i) for i in test]))