🤖 DataTalk - A Multimodal Q&A Chatbot

A powerful AI chatbot that answers questions from documents (PDF, DOCX, CSV), images, and videos using cutting-edge NLP and computer vision models.

Built with Streamlit, Hugging Face Transformers, BLIP, and FAISS.

---

📌 Features

• 🔍 Semantic search using SentenceTransformers + FAISS
• 📄 Text QA from PDF, DOCX, CSV using BERT
• 📊 Handles structured data (tables in DOCX and CSV)
• 🖼️ Visual Q&A using BLIP (Salesforce)
• 🎞️ Video QA by extracting frames using OpenCV
• 🧠 User-friendly Streamlit UI

---

📂 Supported File Types

Type	Formats
Documents	.pdf, .docx, .csv
Images	.jpg, .jpeg, .png
Videos	.mp4, .avi, .mov

---

🧠 Models Used

Task	Model Name
Sentence Embedding	all-MiniLM-L6-v2 (SentenceTransformers)
Text Q&A	bert-large-uncased-whole-word-masking-finetuned-squad
Visual Q&A	Salesforce/blip-vqa-base

---

Output

👨‍💻 Author

Made by Dharshini

---