convert.py

"""
This script converts the data from the raw data to CSV files.

Usage:
    make newsSpace
    python convert.py
"""

import csv
import html
import os
import sys

import pandas as pd

from bs4 import BeautifulSoup

from sklearn.model_selection import train_test_split

HEADER = [
    "source",
    "url",
    "title",
    "image",
    "category",
    "description",
    "rank",
    "pubdate",
]

OUTPUT_FILE = "ag_news.csv"
TRAIN_OUTPUT_FILE = "train.csv"
TEST_OUTPUT_FILE = "test.csv"


def _clean_text(text):
    text = text.replace("\\\n", "\n")
    text = html.unescape(text)

    if text == "\\N":
        return ""

    return text


def _clean_html(text):
    html_code = _clean_text(text)
    html_code.replace("</p>", "\n\n</p>")
    html_code.replace("<br>", "\n")

    soup = BeautifulSoup(html_code, "html.parser")
    text = soup.get_text(separator=" ")

    text = text.replace(" \n", "\n").replace("\n ", "\n")

    # remove extra spaces at the beginning of the text
    lines = [line.strip() for line in text.split("\n")]

    return "\n".join(lines)


def _clean_image(image):
    if image == "none":
        return None
    return image


def _clean_rank(rank):
    return int(rank)


def run():
    """
    Run the conversion process.
    """
    rows = []
    categories = set()

    with open("newsSpace", encoding="ISO-8859-15") as f:
        doc = f.read()

    for row in doc.split("\t\\N\n"):
        if not row:
            continue

        row = row.replace("\\\t", "")

        try:
            source, url, title, image, category, description, rank, pubdate = row.split(
                "\t"
            )
        except ValueError:
            print(repr(row))
            sys.exit(1)

        categories.add(category)

        obj = {
            "source": source,
            "url": url,
            "title": _clean_text(title),
            "image": _clean_image(image),
            "category": category,
            "description": _clean_text(description),
            "rank": _clean_rank(rank),
            "pubdate": pubdate,
            "text": _clean_html(description),
        }

        rows.append(obj)

    # Add a label to each row
    _categories = list(categories)
    _categories.sort()

    save_categories(_categories)

    for row in rows:
        row["label"] = _categories.index(row["category"])

    save_csv(rows)

    split_csv_train_test(test_size=0.2, random_state=42)
    filter_and_split_csv(["World", "Sports", "Business", "Sci/Tech"], "top4")


def save_csv(rows, fname=OUTPUT_FILE):
    """
    Save the processed data into a CSV file.
    """
    with open(fname, "w", encoding="utf8") as f:
        writer = csv.DictWriter(f, fieldnames=rows[0].keys())
        writer.writeheader()

        for row in rows:
            writer.writerow(row)


def split_csv_train_test(**kwargs):
    """
    Split the data into training and testing sets.
    """
    df = pd.read_csv(OUTPUT_FILE)
    train_df, test_df = train_test_split(df, **kwargs)

    os.makedirs("data/all", exist_ok=True)

    train_df.to_csv(os.path.join("data", "all", TRAIN_OUTPUT_FILE), index=False)
    test_df.to_csv(os.path.join("data", "all", TEST_OUTPUT_FILE), index=False)


def filter_and_split_csv(categories, config_name, is_balanced=True, **kwargs):
    """
    Filter the data by categories and split the data into training and testing
    sets. If is_balanced is True, the data will be balanced to size of the
    class with fewer examples.
    """
    df = pd.read_csv(OUTPUT_FILE)

    if is_balanced:
        dfs = []
        for category in categories:
            _df = df[df["category"] == category]
            dfs.append(_df)

        min_size = min([len(_df) for _df in dfs])

        dfs = [df.sample(min_size) for df in dfs]
        df = pd.concat(dfs)
    else:
        df = df[df["category"].isin(categories)]

    os.makedirs(f"data/{config_name}", exist_ok=True)

    train_df, test_df = train_test_split(df, **kwargs)
    train_df.to_csv(os.path.join("data", config_name, TRAIN_OUTPUT_FILE), index=False)
    test_df.to_csv(os.path.join("data", config_name, TEST_OUTPUT_FILE), index=False)


def save_categories(categories, fname="categories.txt"):
    """
    Save the categories into a text file.
    """
    with open(fname, "w") as f:
        for category in categories:
            f.write(category + os.linesep)


if __name__ == "__main__":
    run()