ProfessorAbraham/amharic-streamlit-qa

# 🇪🇹 Amharic QA System using Streamlit This project is a simple Question Answering (QA) system built for Amharic-language documents. It uses sentence-transformer embeddings and a semantic search interface powered by Streamlit. > ✅ Ask any question in Amharic and retrieve the most relevant passages from pre-embedded legal documents. --- ## 🚀 Live Demo 👉 Launch the Streamlit App --- ## 📦 Features - 🔍 **Multilingual Embedding** with `paraphrase-multilingual-mpnet-base-v2` - 📚 **Amharic Legal Chunk Retrieval** from preprocessed `.jsonl` chunks - ⚡ **Semantic Search** using cosine similarity on embedded vectors - 🎈 **Fast & Lightweight UI** using Streamlit --- ## 📁 File Structure # 🇪🇹 Amharic QA System using Streamlit This project is a simple Question Answering (QA) system built for Amharic-language documents. It uses sentence-transformer embeddings and a semantic search interface powered by Streamlit. > ✅ Ask any question in Amharic and retrieve the most relevant passages from pre-embedded legal documents. --- ## 🚀 Live Demo 👉 Launch the Streamlit App *(replace with your live URL once deployed)* --- ## 📦 Features - 🔍 **Multilingual Embedding** with `paraphrase-multilingual-mpnet-base-v2` - 📚 **Amharic Legal Chunk Retrieval** from preprocessed `.jsonl` chunks - ⚡ **Semantic Search** using cosine similarity on embedded vectors - 🎈 **Fast & Lightweight UI** using Streamlit --- --- ## 🧠 Model & Data - **Embedding Model**: `paraphrase-multilingual-mpnet-base-v2` - **Chunked Dataset**: Legal document chunks from Amharic texts - `amharic-chunks-with-embeddings.jsonl` – Google Drive hosted - `amharic-sentences-points.pkl` – precomputed sentence embeddings Files are auto-downloaded in the app via `gdown`. --- ## 🛠️ Installation (Local) ```bash git clone github.com cd amharic-streamlit-qa pip install -r requirements.txt streamlit run app.py