sagayorammbindigili/tanzania-job-credibility-ai

AI-powered system for detecting fake and real online job advertisements in Tanzania using NLP, mBERT, OCR, and Explainable AI. # Development of an Intelligent System for Assessing the Credibility of Online Job Advertisements in Tanzania ## Overview This project focuses on the development of an intelligent web-based system for assessing the credibility of online job advertisements in Tanzania. The system uses Natural Language Processing (NLP), Machine Learning, and Explainable Artificial Intelligence (XAI) techniques to classify online job advertisements as either **Real** or **Fake**. The system allows users to provide job advertisements through: - Text input (copy and paste job descriptions) - Image upload (job advertisement screenshots) For image-based advertisements, Optical Character Recognition (OCR) is applied to extract text before classification. The system supports: - English job advertisements - Kiswahili job advertisements - Mixed English-Kiswahili advertisements --- # Key Features ## 1. Job Advertisement Classification The system classifies advertisements into: - Real Job Advertisement - Fake Job Advertisement and provides a confidence score for the prediction. --- ## 2. Image Advertisement Processing Users can upload images containing job advertisements. The system uses: - EasyOCR for text extraction - mBERT for multilingual text representation - Machine Learning classifier for prediction --- ## 3. Multilingual Support The system is designed for the Tanzanian environment and supports: - English - Kiswahili - Code-switched advertisements --- ## 4. Explainable Prediction The system uses LIME (Local Interpretable Model-Agnostic Explanations) to explain the reasons behind each prediction. Users can understand important factors that influenced the classification result. --- # System Architecture User | | +----------------+ | | Text Input Image Upload | | | ↓ | EasyOCR | | +----------------+ | ↓ Text Preprocessing | ↓ mBERT Model | ↓ Feature Embeddings | ↓ Logistic Regression | ↓ Real / Fake Prediction | ↓ LIME Explanation --- # Technologies Used ## Ba …