Analyzing The October 7, 2023 War Events Based on Telegram Channels

A framework for extracting, processing, and analyzing Telegram communications and related event data around the October 7, 2023 conflict.

This project combines three key modules: message extraction, event parsing, and sentiment analysis, to map sentiment dynamics onto real‐world conflict timelines.

---

Project Structure

. 
├── telegram_groups_messages/       # Telegram message extraction module 
├── wikipedia_events/               # Wikipedia conflict periods parsing module 
└── sentiment_analysis/             # Sentiment analysis & time-series notebooks 

---

Modules

1. Telegram Message Extraction

Fetches raw messages from selected Telegram groups and stores them in SQLite for downstream analysis.

• Location: telegram_groups_messages/
• Core scripts: main.py, telegram_extractor.py, messages_database.py
• Highlights:
  • Rate-limit handling, timezone normalization (Asia/Jerusalem), media metadata, message-level metrics.
  • Composite-key SQLite schema with JSON serialization for complex fields.
• See module README.

2. Wikipedia Events Parsing

Builds a structured timeline of conflict phases by extracting dates and section titles from Wikipedia.

• Location: wikipedia_events/
• Core scripts: wikipedia_dates_parser.py, wikipedia_titles_parser.py, time_periods_parser.py
• Highlights:
  • HTML scraping and API calls to capture diverse date formats.
  • Maps section headings to (start_date, end_date, duration) records for correlation.
• See module README.

3. Sentiment Analysis & Time Series

Applies multilingual sentiment models to the Telegram dataset, then aggregates, decomposes, and detects events in the sentiment series.

• Location: sentiment_analysis/
• Core notebooks:
  1. sentiment_analysis_1.ipynb – Message-level sentiment labeling
  2. sentiment_agg_by_time_2.ipynb – Hourly/daily/weekly aggregation
  3. sentiment_time_series_3.ipynb – Decomposition & stationarity tests
  4. sentiment_time_series_by_type_4.ipynb – Group-type comparisons & Granger tests
  5. sentiment_event_detection_5.ipynb – Change-point & cross-correlation detection
  6. sentiment_events_analysis_6.ipynb – Event correlation & statistical validation
• Highlights:
  • Hebrew (DicataBERT-Sentiment) and Arabic (CAMeLBERT-DA) classification.
  • Net Sentiment & Sentiment Index metrics, CUSUM/PELT detection, t-tests/ANOVA, regression models.
• See module README.

---

Prerequisites

• Python ≥ 3.11
• SQLite (built-in)
• Jupyter Notebook/Lab
• Common packages:

  pip install telethon tqdm wakepy python-dotenv emoji pytz requests beautifulsoup4 wikipedia-api nltk pandas numpy matplotlib nltk transformers torch statsmodels ruptures scipy plotly

  python -m nltk.downloader punkt

---

Usage Workflow

1. Extract Telegram messages

cd telegram_groups_messages
python main.py

2. Parse Wikipedia conflict periods

cd ../wikipedia_events
python wikipedia_titles_parser.py
python wikipedia_dates_parser.py
python time_periods_parser.py

3. Run sentiment analysis pipeline

cd ../sentiment_analysis
jupyter nbconvert --to notebook --execute sentiment_analysis_1.ipynb
# repeat for each notebook in order 2 → 6

---

Customization

• Time windows & thresholds: adjust in main.py and analysis notebooks
• Date formats & section depths: extend regexes or recursion levels in parsers
• Models & metrics: swap sentiment classifiers or add forecasting