RedBullet12/kazakh-rag-qa

🇰🇿 Kazakh RAG Question Answering

A Retrieval-Augmented Generation (RAG) system for Kazakh-language QA.

How it works

1. 1.Your question is embedded using KazEmbed-v5
2. 2.FAISS retrieves the top-K most similar passages from KazQAD
3. 3.RoBERTa extracts the answer span from the best passage

Dataset

KazQAD by ISSAI, Nazarbayev University

Stack

• —sentence-transformers + Nurlykhan/kazembed-v5
• —faiss-cpu
• —deepset/roberta-base-squad2
• —Gradio