Update README.md

README.md

@@ -1,146 +1,159 @@
-📄 AI Research Paper Explainer
-
-Hugging Face · RAG · CPU-only · No Paid APIs
-
-An end-to-end Retrieval-Augmented Generation (RAG) system that reads long research papers and explains them in simple, grounded terms.
-
-Built to deeply understand how embeddings, retrieval, and generation work together — not just to make another chatbot.
-
-🚀 What This Project Does
-
-Given a text-based research paper (PDF), the system:
-
-Extracts and processes long documents safely
-
-Chunks content into semantically meaningful pieces
-
-Builds embeddings using Sentence-Transformers
-
-Retrieves the most relevant sections via cosine similarity
-
-Generates grounded explanations using a Hugging Face LLM
-
-Supported Analysis Modes
-
-Summary – Explain the paper in simple terms
-
-Main Contribution – What is new or novel
-
-Why Better – Improvements over prior work
-
-Limitations – Weaknesses or constraints
-
-Prerequisites – Concepts to learn before reading
-
-🧠 Architecture (High-Level)
-PDF (text-based)
-   ↓
-Text Extraction
-   ↓
-Chunking (long-document safe)
-   ↓
-Embeddings (Sentence-Transformers)
-   ↓
-Semantic Similarity (Top-K Retrieval)
-   ↓
-Context Builder
-   ↓
-Hugging Face LLM (FLAN-T5)
-   ↓
-Grounded Explanation
-
-⚠️ Scanned PDFs are intentionally excluded to ensure fast, reliable, free-tier-safe processing.
-
-⚠️ Scanned PDFs are intentionally excluded to ensure fast, reliable, free-tier-safe processing.
-
-📂 Project Structure
-Research/
-│
-├── app.py                     # CLI pipeline runner
-├── streamlit_app.py           # Streamlit UI
-│
-├── extractor/
-│   └── pdf_parser.py
-│
-├── chunker/
-│   └── text_chunker.py
-│
-├── embeddings/
-│   ├── embedder.py
-│   └── similarity.py
-│
-├── llm/
-│   ├── explainer.py
-│   └── question_bank.py
-│
-├── utils/
-│   ├── validators.py
-│   └── context_builder.py
-│
-├── requirements.txt
-└── README.md
-
-⚙️ Setup & Installation
-1️⃣ Create virtual environment
-python -m venv venv
-source venv/bin/activate   # Linux/Mac
-venv\Scripts\activate      # Windows
-2️⃣ Install dependencies
-pip install -r requirements.txt
-
-3️⃣ Run Streamlit app
-streamlit run streamlit_app.py
-
-
-Upload a text-based research paper (PDF) and select an analysis mode.
-
-🧪 Example Output
-
-Mode: Summary
-
-This paper introduces the Transformer architecture, which replaces recurrent networks with self-attention mechanisms. This allows parallel processing of sequences, leading to faster training and better performance on sequence-to-sequence tasks.
-
-🎯 Why This Project Matters
-
-Most GenAI demos:
-
-send full documents to an LLM
-
-ignore retrieval
-
-hallucinate confidently
-
-This project:
-
-controls context explicitly
-
-retrieves before generating
-
-mirrors how real RAG systems are built in production
-
-📌 Design Decisions
-
-❌ No OCR → avoids heavy dependencies and free-tier crashes
-
-✅ Text-based PDFs only (arXiv-style)
-
-✅ Chunked processing for long documents
-
-✅ Top-K retrieval to reduce hallucination
-
-These choices prioritize robustness, clarity, and deployability.
-
-🔮 Future Improvements
-
-Section-aware chunking (Abstract / Method / Results)
-
-Hugging Face Spaces deployment
-
-Evaluation metrics for retrieval quality
-
-Lightweight vector store (FAISS)
-
-🙌 Author
-
-Built by Adarsh
+---
+title: AI Research Paper Explainer
+emoji: 📄
+colorFrom: blue
+colorTo: purple
+sdk: streamlit
+sdk_version: "1.31.0"
+python_version: "3.10"
+app_file: streamlit_app.py
+pinned: false
+---
+
+
+📄 AI Research Paper Explainer
+
+Hugging Face · RAG · CPU-only · No Paid APIs
+
+An end-to-end Retrieval-Augmented Generation (RAG) system that reads long research papers and explains them in simple, grounded terms.
+
+Built to deeply understand how embeddings, retrieval, and generation work together — not just to make another chatbot.
+
+🚀 What This Project Does
+
+Given a text-based research paper (PDF), the system:
+
+Extracts and processes long documents safely
+
+Chunks content into semantically meaningful pieces
+
+Builds embeddings using Sentence-Transformers
+
+Retrieves the most relevant sections via cosine similarity
+
+Generates grounded explanations using a Hugging Face LLM
+
+Supported Analysis Modes
+
+Summary – Explain the paper in simple terms
+
+Main Contribution – What is new or novel
+
+Why Better – Improvements over prior work
+
+Limitations – Weaknesses or constraints
+
+Prerequisites – Concepts to learn before reading
+
+🧠 Architecture (High-Level)
+PDF (text-based)
+   ↓
+Text Extraction
+   ↓
+Chunking (long-document safe)
+   ↓
+Embeddings (Sentence-Transformers)
+   ↓
+Semantic Similarity (Top-K Retrieval)
+   ↓
+Context Builder
+   ↓
+Hugging Face LLM (FLAN-T5)
+   ↓
+Grounded Explanation
+
+⚠️ Scanned PDFs are intentionally excluded to ensure fast, reliable, free-tier-safe processing.
+
+⚠️ Scanned PDFs are intentionally excluded to ensure fast, reliable, free-tier-safe processing.
+
+📂 Project Structure
+Research/
+│
+├── app.py                     # CLI pipeline runner
+├── streamlit_app.py           # Streamlit UI
+│
+├── extractor/
+│   └── pdf_parser.py
+│
+├── chunker/
+│   └── text_chunker.py
+│
+├── embeddings/
+│   ├── embedder.py
+│   └── similarity.py
+│
+├── llm/
+│   ├── explainer.py
+│   └── question_bank.py
+│
+├── utils/
+│   ├── validators.py
+│   └── context_builder.py
+│
+├── requirements.txt
+└── README.md
+
+⚙️ Setup & Installation
+1️⃣ Create virtual environment
+python -m venv venv
+source venv/bin/activate   # Linux/Mac
+venv\Scripts\activate      # Windows
+2️⃣ Install dependencies
+pip install -r requirements.txt
+
+3️⃣ Run Streamlit app
+streamlit run streamlit_app.py
+
+
+Upload a text-based research paper (PDF) and select an analysis mode.
+
+🧪 Example Output
+
+Mode: Summary
+
+This paper introduces the Transformer architecture, which replaces recurrent networks with self-attention mechanisms. This allows parallel processing of sequences, leading to faster training and better performance on sequence-to-sequence tasks.
+
+🎯 Why This Project Matters
+
+Most GenAI demos:
+
+send full documents to an LLM
+
+ignore retrieval
+
+hallucinate confidently
+
+This project:
+
+controls context explicitly
+
+retrieves before generating
+
+mirrors how real RAG systems are built in production
+
+📌 Design Decisions
+
+❌ No OCR → avoids heavy dependencies and free-tier crashes
+
+✅ Text-based PDFs only (arXiv-style)
+
+✅ Chunked processing for long documents
+
+✅ Top-K retrieval to reduce hallucination
+
+These choices prioritize robustness, clarity, and deployability.
+
+🔮 Future Improvements
+
+Section-aware chunking (Abstract / Method / Results)
+
+Hugging Face Spaces deployment
+
+Evaluation metrics for retrieval quality
+
+Lightweight vector store (FAISS)
+
+🙌 Author
+
+Built by Adarsh
 Learning GenAI by building systems, not just prompts.