Ghena-A/Research-to-social-media-Generator

R2S is an AI-powered tool that transforms research content (PDFs or summaries) into platform-specific social media posts. It applies text preprocessing, summarization, and prompt engineering to ensure clear, professional outputs in Arabic and English. Built as a final AI course project. # Research-to-social-media-Generator R2S is an AI-powered tool that transforms research content (PDFs or summaries) into platform-specific social media posts. It applies text preprocessing, summarization, and prompt engineering to ensure clear, professional outputs in Arabic and English. Built as a final AI course project. # Research to Social Media Generator (R2S) R2S is an AI-powered application that transforms research content (PDFs or written summaries) into high-quality, platform-specific social media posts. The project focuses on **bridging the gap between academic research and social media communication**, ensuring that complex ideas are rewritten in a clear, professional, and audience-appropriate way. --- ## 🚀 Features - Upload **one research PDF** or paste a **written summary** - Automatic **text extraction, cleaning, and preprocessing** - AI-powered **summarization and content restructuring** - Platform-specific post generation for: - LinkedIn - Instagram - Facebook - Supports **Arabic and English** - Ensures: - Clear structure - Professional tone - No marketing or promotional hallucinations - Platform-appropriate style - Clean UI built with **Streamlit** --- ## 🧠 AI Pipeline Overview 1. **PDF Processing** - Extract text using PyPDF2 - Sentence filtering and noise removal - Chunking long documents for safe processing 2. **Summarization** - Uses a transformer-based summarization model - Produces concise, meaningful research summaries 3. **Prompt Engineering** - Carefully designed prompts per platform and language - Ensures consistent tone and structure - Prevents unwanted outputs (ads, services, irrelevant content) 4. **LLM Generation** - Supports OpenRouter and Ollama - Automatically falls back if one provider fails 5. **Post-processing** - Arabic RTL fixes - Token cleanup - English word filtering for Arabic platforms - Hashtag generation --- ## 🛠 Tech Stack - Python - Streamlit - Transformers (HuggingFace) - PyPDF2 - OpenRouter API - Ollama …