extract_news/3_extract_news_txt.py

# -*- coding: utf-8 -*-

"""
News article text extractor
 
This script extracts the text from locally-stored news articles. The main goal is 
to retrieve clean text with minimal external elements, such as user menus, article lists, 
and advertisements.
 
To install the necessary packages:
pip install mysql-connector-python chardet colorama pyquery
 
After completing this step, you can use the Python script located at /dataset/2_cleaning_txt.py 
to standardize the text for your dataset.
  
Note:
RSS feed links for media sources, as well as the HTML structure of media pages, tend to change 
and evolve regularly. It's crucial to regularly check the output per media source and adjust 
the parsing process to ensure high-quality text extraction and to address potential changes 
in RSS feed URLs.

Author     : Guillaume Eckendoerffer
Date       : 29-09-23
Repository : https://github.com/Eckendoerffer/TorchTrainerFlow/
             https://huggingface.co/datasets/eckendoerffer/news_fr
"""

import chardet
import time, os, re, html, json, hashlib
from colorama import Fore, init
from pyquery import PyQuery as pq
from config import DB_CONFIG
from utils import create_connection, get_file_content, save_to_file, clean_text, decode_unicode_escapes, decode_content

index_id  = 1 
stop_id   = 1000000
path = os.getcwd()
init(autoreset=True)

connection = create_connection(DB_CONFIG)
cursor = connection.cursor()
query = "SELECT `key_title` FROM `base_news` WHERE `key_title` != ''"
cursor.execute(query)
keys = cursor.fetchall()
formatted_keys = "|".join([key[0] for key in keys]) + "|"

while True:
    id_source =''
    next_id = index_id + 1
    time_start = time.time()
    with connection.cursor() as cursor:
        cursor = connection.cursor(dictionary=True)
        cursor.execute(f"SELECT `id`, `url`, `media`, `key_title` FROM `base_news` WHERE `id`='{index_id}' LIMIT 1")
        row = cursor.fetchone()
        if row:
            id_source = row["id"]
            id_media = row["media"]
            key_title = row["key_title"]
            url = row["url"].strip()

    if key_title.strip():
        index_id = next_id
        continue
    
    if id_source and id_source > stop_id:
        break

    # Retrieve HTML content from the file    
    html_content = ''
    content=''
    title=''
    file_path = os.path.join(path, "sources", "html_news", f"{id_source}.txt")
    html_content = get_file_content(file_path)

    if '/replay' in url or not html_content:
        index_id = next_id
        continue
 
    len_source = len(html_content)

    # encoding
    if isinstance(html_content, str):
        html_content_bytes = html_content.encode('utf-8')
    else:
        html_content_bytes = html_content
    decoded_content = decode_content(html_content_bytes)
    if decoded_content is None:
        charset_result = chardet.detect(html_content_bytes)
        current_encoding = charset_result['encoding']
        try:
            html_content = html_content_bytes.decode(current_encoding)
        except Exception as e:
            print(Fore.WHITE + f"Error: {e}")
            index_id = next_id
            continue
    else:
        html_content = decoded_content

    len_or = len(html_content)

    # Use pyquery to parse the HTML
    try:
        doc = pq(html_content)
    except Exception as e: 
        print(Fore.WHITE + f"({id_source}) Error parsing HTML: {e} {url}")
        index_id = next_id
        continue

    # Extracting the title
    if title.strip() =='':
        title = html.unescape(doc('h1:first').text())
        if title.strip() =='': 
            index_id = next_id
            continue

    extract_method = 0
    # Extracting the text from "articleBody": ld json
    match = re.search(r'"articleBody"\s*:\s*"([^"]+)"', html_content)
    if match:
        content = html.unescape(match.group(1))
        if content.strip(): 
            extract_method = 1
            try:
                content = json.loads(f'"{content}"')
            except json.JSONDecodeError:
                content = decode_unicode_escapes(content)

    # Extracting the text from <article> tag
    if not extract_method or len(content) < 100:
        p_elements = doc('article p')
        if not p_elements:
            p_elements = doc('div.Body p')
        if not p_elements:
            p_elements = doc('div.post-body p')
        if not p_elements:
            p_elements = doc('div.article_content p')
        if not p_elements:
            p_elements = doc('div.article__text p')
        if not p_elements:
            p_elements = doc('div.article-description p')
        if not p_elements:
            p_elements = doc('div.mainBody p')
        if not p_elements:
            p_elements = doc('section.article-section p')
        if not p_elements:
            p_elements = doc('div.article p')
            
        for p in p_elements:
            html_element = pq(p)
            html_content += f" {html_element.html().strip()} "
        
            if ".futura-sciences.com" in url:
                html_element.find('a').remove()
            content += f" {html.unescape(html_element.text().strip())} "
        if content.strip(): extract_method = 2

    len_text = len(content)

    # Adding a space after punctuation and various deletions
    content = content.replace('\r', ' ').replace('\n', ' ')
    remove_phrases = ['.', '?', '!', ';', '!', '»', ']']
    for phrase in remove_phrases:
        content = content.replace(phrase, phrase + ' ')
    content = content.replace(html.unescape('&nbsp;'), ' ')
    content = re.sub(r'\s{2,}', ' ', content)
    content = re.sub(r'À lire aussi.{1,200}?»', ' ', content)
    content = re.sub(r'Temps de Lecture.{1,20}? Fiche', ' ', content)
    content = re.sub(r'Temps de Lecture.{1,20}? min.', ' ', content)

    # Media-specific cutting rules
    if ".elle.fr" in url and "sur 5" in content:
        content = re.sub(r'Note :.{13,45}? sur 5', ' ', content)
        content = content.replace(content[content.find(" sur 5 "):], ' ')

    if ".latribune.fr" in url:
        content = content.replace("Partager :", ' ')

    removePhrasesEnd = [
        'Sur le même sujet',
        'Sur lemême thème',
        'Nos articles à lire aussi',
        'Suivez toute l’actualité de vos villes',
        'En direct',
        "J'ai déjà un compte",
        '> Ecoutez',
        "Lire aussi >>",
        'Courrier international',
        'Vous avez trouvé une erreur?',
        'Il vous reste',
        'Partager',
        "Suivez-nous",
        'Newsletter',
        'Abonnez-vous',
        '1€ le premier mois',
        'Votre France Bleu',
        'Soyez le premier à commenter cet article',
        'Pour rester informé(e)',
        'Un site du groupe',
        "Cet article est réservé aux abonnés",
        "Recevez chaque vendredi l'essentiel",
        "Suivez toute l'actualité de ZDNet",
        "Suivez-nous sur les résaux sociaux",
        ". par ",
        "Le résumé de la semaine",
        "ACTUELLEMENT EN KIOSQUE",
        " L’actualité par la rédaction de",
        "Gratis onbeperkt",
        "Débloquez immédiatement cet article",
        "À voir également",
        "null null null ",
        'Du lundi au vendredi, à 19h',
        "La rédaction de La Tribune",
        "Restez toujours informé: suivez-nous sur Google Actualités",
        "Du lundi au vendredi, votre rendez-vous",
        "Enregistrer mon nom, mon e-mail",
        "Mot de passe oublié",
        "accès à ce contenu",
        "En cliquant sur",
        '(function',
    ]
    # Remove sentences at the end
    for phrase in removePhrasesEnd:
        if phrase in content:
            content = content.split(phrase, 1)[0]

    removePhrases = [
        "Inscrivez-vous pour recevoir les newsletters de la Rép' dans votre boîte mail",
        "TF1 INFO",
        "Sujet TF1 Info",
        "Sujet JT LCI",
        "TF1 Info ",
        "JT 20h WE ",
        "JT 20h Semaine ",
        "Source :",
        "Inscrivez-vous aux newsletters de la RTBF Tous les sujets de l'article",
        "Pour voir ce contenu, connectez-vous gratuitement",
        ">> LIRE AUSSI",
        "À LIRE AUSSI",
        "A lire aussi >> ",
        "» LIRE AUSSI -",
        " → À LIRE.",
        "À voir également",
        "Image d'illustration -",
        "Le média de la vie locale ",
        "Les plus lus.",
        "Ce live est à présent terminé.",
        " . -",
        "[…]",
        "[.]",
        "(…)",
        "(.)",
        "©",
        "Tous droits réservés",
        " sur TF1",
        "Avec AFP",
        " AFP /",
        "/ AFP ",
        ". AFP",
        " BELGA /",
        "GETTY",
        "Getty Images",
        "→ EXPLICATION",
        "→ LES FAITS",
        "→",
        "À lire aussi",
        "EN RÉSUMÉ",
        "•",
        "►►► ",
        "► Écoutez l'entièreté de ce podcast ci-dessus",
        "► Pour écouter ce podcast , abonnez-vous",
        "►"
    ]

    # Remove terms
    for phrase in removePhrases:
        content = content.replace(phrase, '.')

    # Replace sequences of characters
    content = content.replace(',.', '.')
    content = content.replace('.,', '.')
    content = re.sub(r'\.{2,}', '.', content)
    content = re.sub(r'\s{2,}', ' ', content)
    content = re.sub(r'-{2,}', '-', content)
    content = re.sub(r'__{2,}', '_', content)
    content = re.sub(r'Publié le\s?:? \d{2} ?\/ ?\d{2} ?\/ ?\d{4}( à \d{2}h\d{2})?', '', content)
    content = re.sub(r'Mis à jour le \d{1,2} \w+ \. \d{4}', '', content)
    matches = [match.group() for match in re.finditer(r'(\d{2}:\d{2})?Modifié le : \d{2}/\d{2}/\d{4}( - \d{2}:\d{2})?', content) if len(match.group()) <= 38]
    for match in matches:
        content = content.replace(match, ' ')

    # Format the content for the txt file into the 'add' variable
    content = re.sub(r'<.*?>', '', content)  
    add = f"{title}. "
    if len(content) > 160:
        add += f"{content}."

    add = clean_text(add)
    key = hashlib.md5(title.encode()).hexdigest()
    nb_base = formatted_keys.count(f'{key}|')

    # Display
    color = Fore.GREEN if len(content) > 200 else Fore.WHITE
    if len(content) > 200 and nb_base: 
        color = Fore.CYAN

    if len(content) > 200 and not nb_base:
        cursor = connection.cursor()
        cursor.execute("UPDATE `base_news` SET `key_title`=%s WHERE `id`=%s LIMIT 1", (key, id_source))
        formatted_keys = f'{formatted_keys}{key}|'
        save_to_file(f"{path}/sources/txt_news/{id_source}.txt", add)

    elapsed_time = time.time() - time_start
    print(color + f"{id_source:8}) ({extract_method:1}) [{elapsed_time:.3f}] [{len_source:7}{len_text:7}{len(content):7}{len(title):4} ] {url} ")
    index_id = next_id