nat_datasets.py

import tqdm
import re
import requests
from huggingface_hub import login
from datasets import Dataset
from datasets import load_dataset
from google.colab import userdata

login(userdata.get('HF_TOKEN'))

data = load_dataset("Ibrahemqasim/categories_en2ar", split="train")

# nationalities.keys() "nat_en","man","men","women","womens","country_en","country_ar",
nationalities = load_dataset("Ibrahemqasim/nationalities", split="train")
nationalities_pattern = r'\b(' + '|'.join(map(re.escape, [n["nat_en"] for n in sorted(nationalities, key=lambda x: -x["nat_en"].count(' '))])) + r')\b'

countries = load_dataset("Ibrahemqasim/countries", split="train")
countries_pattern = r'\b(' + '|'.join(map(re.escape, [n["en"] for n in sorted(countries, key=lambda x: -x["en"].count(' '))])) + r')\b'

# ---
countries_dict = {cc["en"]: cc for cc in countries}
nationalities_dict = {cc["nat_en"]: cc for cc in nationalities}
# ---
to_work = [
    "categories_with_nationalities",
    # "categories_with_years",
    # "categories_with_YEAR_COUNTRY_pattern",
    # "categories_with_YEAR_pattern",
]

data_lists = {
    "categories_with_nationalities" : {},
    "categories_with_years" : {},
    "categories_with_YEAR_COUNTRY_pattern" : {},
    "categories_with_YEAR_pattern" : {},
}

YEAR_PATTERN = "{YEAR}"
NAT = "{NAT}"
AR_NAT_MEN = "{NAT_MEN}"
COUNTRY_PATTERN = "{COUNTRY}"

for tab in tqdm.tqdm(data):
    # ---
    key = tab["en"]
    value = tab["ar"]
    # ---
    # Add if key and value has 4 digits and they are the same
    reg_year = r"(\d+[–-]\d+|\d{4})"
    # ---
    key_digits = re.search(reg_year, key)
    value_digits = re.search(reg_year, value)
    # ----
    match1 = re.search(nationalities_pattern, key)
    # ----
    if match1:
        en_country = match1.group(1)
        ar_country = nationalities_dict.get(en_country, {}).get("men", "")
        # ---
        if ar_country and ar_country in value:
            key1 = re.sub(rf'\b{re.escape(en_country)}\b', COUNTRY_PATTERN, key)
            value1 = re.sub(rf'\b{re.escape(ar_country)}\b', AR_NAT_MEN, value)
            # ---
            if COUNTRY_PATTERN in key1 and AR_NAT_MEN in value1:
                # ---
                if key1 in data_lists["categories_with_nationalities"]:
                    data_lists["categories_with_nationalities"][key1]["count"] += 1
                else:
                    data_lists["categories_with_nationalities"][key1] = {"ar": value1, "count": 1}
    # ---
    continue
    # ---
    if key_digits and value_digits and key_digits.group() == value_digits.group():
        # data_lists["categories_with_years"].append({"en": key, "ar": value, "count": 1})
        if key in data_lists["categories_with_years"]:
            data_lists["categories_with_years"][key]["count"] += 1
        else:
            data_lists["categories_with_years"][key] = {"ar": value, "count": 1}
        # ---
        key2 = key.replace(key_digits.group(), YEAR_PATTERN)
        value2 = value.replace(value_digits.group(), YEAR_PATTERN)
        # ---
        # data_lists["categories_with_YEAR_pattern"].append({"en": key2, "ar": value2, "count": 1})
        # ---
        if key2 in data_lists["categories_with_YEAR_pattern"]:
            data_lists["categories_with_YEAR_pattern"][key2]["count"] += 1
        else:
            data_lists["categories_with_YEAR_pattern"][key2] = {"ar": value2, "count": 1}
        # ----
        # البحث عن اسم الدولة في key2
        match = re.search(countries_pattern, key2)
        # ----
        if match:
            en_country = match.group(1)
            ar_country = countries.get(en_country)
            # ---
            if ar_country and ar_country in value2:
                key3 = re.sub(rf'\b{re.escape(en_country)}\b', COUNTRY_PATTERN, key2)
                value3 = re.sub(rf'\b{re.escape(ar_country)}\b', COUNTRY_PATTERN, value2)
                # ---
                if COUNTRY_PATTERN in key3 and COUNTRY_PATTERN in value3:
                    # ---
                    if key3 in data_lists["categories_with_YEAR_COUNTRY_pattern"]:
                        data_lists["categories_with_YEAR_COUNTRY_pattern"][key3]["count"] += 1
                    else:
                        data_lists["categories_with_YEAR_COUNTRY_pattern"][key3] = {"ar": value3, "count": 1}
        # ----
# ----
print(f"{len(data_lists['categories_with_YEAR_COUNTRY_pattern'])=}")
print(f"{len(data_lists['categories_with_YEAR_pattern'])=}")

print(f"all data len: {len(data):,}.")

# for x, data_list in data_lists.items():
for x in to_work:
    data_list = data_lists.get(x)
    # ---
    if x == "countries":
        data_list = [{"en": key, "ar": value} for key, value in data_list.items()]
    else:
        data_list = [{"en": key, "ar": value["ar"], "count": value["count"]} for key, value in data_list.items()]
        # sort data_list by count
        data_list = sorted(data_list, key=lambda x: x["count"], reverse=True)
    # ---
    print("______________")
    print(f"len of {x} : {len(data_list)}.")
    # ---
    print("____________________________")
    # ---
    # إنشاء Dataset
    dataset = Dataset.from_list(data_list)

    # رفع Dataset إلى Hugging Face
    dataset.push_to_hub(f"Ibrahemqasim/{x}")
    # ---
    print(f"dataset: Ibrahemqasim/{x} push_to_hub successfully!")