⚖️ LegalLens

AI-powered legal document analysis — upload a contract, get summaries, and ask questions in your language.

---

Overview

LegalLens makes legal documents understandable. Upload a PDF, scanned image, or text file and the system extracts text, translates if needed, summarises long passages, and indexes the content for retrieval. You can then chat with the document in your preferred language and get structured, source-backed answers.

Pipeline

Upload → OCR / pdfplumber → Language detect & translate
       → BigBird-Pegasus summarisation
       → Embeddings → Pinecone (retrieval)
       → User question → m2m100 translate
       → Haystack EmbeddingRetriever → Pinecone
       → Seq2SeqGenerator (BART LFQA) → Answer
       → MongoDB chat history → translate back to user language

Features

Capability	Implementation
Document ingestion	PDFs, scanned images, plain text
OCR	Tesseract.js for scanned legal docs
Translation	Facebook m2m100 (multi-direction)
Summarisation	BigBird-Pegasus with chunking
Vector search	Pinecone (us-west4-gcp-free, cosine, 768d)
QA generation	Haystack Seq2SeqGenerator + vblagoje/bart_lfqa
Context retention	MongoDB conversation history
UI	React/Vue web frontend served via Flask

Tech Stack

• Backend: Flask + Haystack + HuggingFace Transformers
• NLP models: BERT (length tokeniser), BART LFQA (generation), m2m100 (translation), BigBird-Pegasus (summarisation), flax-sentence-embeddings/all_datasets_v3_mpnet-base (embeddings)
• Storage: Pinecone (vectors), MongoDB (chat history)
• Frontend: Web app under webapp/

Getting Started

git clone https://github.com/bhavya-x/LegalLens.git
cd LegalLens
pip install -r requirements.txt   # if present
# Configure Pinecone + MongoDB credentials in includes/dependencies.py
python server.py

Repository Layout

LegalLens/
├── server.py              # Flask app + Haystack pipeline
├── includes/              # Shared dependencies and helpers
├── data/                  # Sample legal datasets
├── models/                # Model artefacts / configs
├── webapp/                # Frontend
├── plots/                 # Architecture diagrams
└── readme.md              # Original project notes

Screenshots

Landing	Summary	Chat