olaifahabeeb-ola/naija-hatescan

# Naija HateScan 🛡️ ### Multilingual Hate Speech Detection System for Nigerian Social Media > **HND Final Year Project** — Federal Polytechnic Offa, Department of Computer Science & Information Technology --- ## Overview **Naija HateScan** is a web-based machine learning system that automatically classifies Nigerian social media posts as **Hate Speech**, **Offensive**, or **Neutral** content. It supports five language varieties: | Language | Notes | |---|---| | English | Nigerian Standard English | | Nigerian Pidgin | Nigerian Creole (code-mixed) | | Yoruba | Southwest Nigeria | | Hausa | Northern Nigeria | | Igbo | Southeast Nigeria | The classifier uses character-level TF-IDF n-gram features with Logistic Regression, making it effective even on code-mixed text where words switch between languages mid-sentence. --- ## Features - **Single-text analysis** with AJAX (no page reload) and inline results - **Batch analysis** — paste up to 100 posts or upload a CSV file - **Severity scoring** — Mild / Moderate / Severe for each harmful post - **Target detection** — Ethnic / Religious / Political / General targeting - **Plain-language explanations** and flagged term highlighting - **Analysis history** with filter, pagination, and per-record delete - **Analytics dashboard** — Chart.js charts for trends, languages, severity - **PDF export** via ReportLab for individual analysis reports - **Auto language detection** with Nigerian variety bias - **No user authentication** — single public-facing tool --- ## Tech Stack | Layer | Technology | |---|---| | Backend | Python 3.10+, Flask 3.x | | Database | SQLite via Flask-SQLAlchemy | | ML | Scikit-learn (TF-IDF + Logistic Regression) | | Language ID | langdetect + custom keyword voting | | Frontend | Bootstrap 5.3, Chart.js 4, Font Awesome 6 | | PDF | ReportLab 4.x | | Model Persistence | joblib | --- ## Project Structure ``` naija_hatescan/ ├── app.py # Flask app, routes, Analysis DB model ├── con …