process/tweet_ner7.py

import json
import os
from random import shuffle, seed
from datasets import load_dataset

test = load_dataset("cardiffnlp/super_tweeteval", "tweet_ner7", split="test").shuffle(seed=42)
test = list(test.to_pandas().T.to_dict().values())
train = load_dataset("cardiffnlp/super_tweeteval", "tweet_ner7", split="train").shuffle(seed=42)
train = list(train.to_pandas().T.to_dict().values())
validation = load_dataset("cardiffnlp/super_tweeteval", "tweet_ner7", split="validation").shuffle(seed=42)
validation = list(validation.to_pandas().T.to_dict().values())
n_train = len(train)
n_validation = len(validation)
for data in [train, validation, test]:
    for i in data:
        i["gold_label_sequence"] = i["gold_label_sequence"].tolist()
        i["entities"] = {k: v.tolist() for k, v in i["entities"].items()}
        i["text_tokenized"] = i["text_tokenized"].tolist()

n_test = int(len(test)/4)
test_1 = test[:n_test]
test_2 = test[n_test:n_test*2]
test_3 = test[n_test*2:n_test*3]
test_4 = test[n_test*3:]

os.makedirs("data/tweet_ner", exist_ok=True)
with open("data/tweet_ner/test.jsonl", "w") as f:
    f.write("\n".join([json.dumps(i) for i in test]))
with open("data/tweet_ner/test_1.jsonl", "w") as f:
    f.write("\n".join([json.dumps(i) for i in test_1]))
with open("data/tweet_ner/test_2.jsonl", "w") as f:
    f.write("\n".join([json.dumps(i) for i in test_2]))
with open("data/tweet_ner/test_3.jsonl", "w") as f:
    f.write("\n".join([json.dumps(i) for i in test_3]))
with open("data/tweet_ner/test_4.jsonl", "w") as f:
    f.write("\n".join([json.dumps(i) for i in test_4]))
with open("data/tweet_ner/train.jsonl", "w") as f:
    f.write("\n".join([json.dumps(i) for i in train]))
with open("data/tweet_ner/validation.jsonl", "w") as f:
    f.write("\n".join([json.dumps(i) for i in validation]))


def sampler(dataset_test, r_seed):
    seed(r_seed)
    shuffle(dataset_test)
    shuffle(train)
    shuffle(validation)
    test_tr = dataset_test[:int(n_train / 2)]
    test_vl = dataset_test[int(n_train / 2): int(n_train / 2) + int(n_validation / 2)]
    new_train = test_tr + train[:n_train - len(test_tr)]
    new_validation = test_vl + validation[:n_validation - len(test_vl)]
    return new_train, new_validation


id2test = {n: t for n, t in enumerate([test_1, test_2, test_3, test_4])}
for n, _test in enumerate([
        test_4 + test_2 + test_3,
        test_1 + test_4 + test_3,
        test_1 + test_2 + test_4,
        test_1 + test_2 + test_3]):
    for s in range(3):
        os.makedirs(f"data/tweet_ner_test{n}_seed{s}", exist_ok=True)
        _train, _valid = sampler(_test, s)
        with open(f"data/tweet_ner_test{n}_seed{s}/train.jsonl", "w") as f:
            f.write("\n".join([json.dumps(i) for i in _train]))
        with open(f"data/tweet_ner_test{n}_seed{s}/validation.jsonl", "w") as f:
            f.write("\n".join([json.dumps(i) for i in _valid]))
        with open(f"data/tweet_ner_test{n}_seed{s}/test.jsonl", "w") as f:
            f.write("\n".join([json.dumps(i) for i in id2test[n]]))