DocuMind OCR App
DocuMind is an intelligent OCR-based Streamlit app that reads scanned documents, corrects typos, extracts keywords, generates concise summaries using NLP models, and translates summaries to English when needed.
Features
- Upload scanned images or PDFs
- Preprocess image for better OCR accuracy
- Clean & format raw OCR text
- Spellcheck and auto-correction
- Extract top keywords
- Generate intelligent summaries using
facebook/bart-large-cnn - Translate non-English summaries to English
- Automatically detects language (using
langdetect)
Tech Stack
| Layer | Tool / Library |
|---|---|
| Frontend | Streamlit |
| OCR Engine | Tesseract OCR (pytesseract) |
| NLP | HuggingFace Transformers (bart, opus-mt) |
| Keyword Ext. | CountVectorizer from Scikit-learn |
| Summarization | facebook/bart-large-cnn |
| Translation | Helsinki-NLP/opus-mt-ROMANCE-en |
| Language Det. | langdetect |
| Preprocessing | OpenCV, PIL, regex |
Screenshots
How to Run Locally
- Clone the repo: ```bash git clone https://github.com/ayulikhar/documind-ocr-app.git cd documind-ocr-app
