🤖 QueryGenie

Offline NLP-Powered Text-to-SQL Engine (NLIDB)

🧠 Built as a 6th Semester NLP Project using classical Natural Language Processing techniques in Python.

Transform natural language into executable SQL — fully offline, privacy-first, and zero API cost.

---

---

🎓 Academic Context

QueryGenie was developed as part of a 6th Semester Natural Language Processing project, with a strong focus on implementing core NLP concepts from scratch using Python rather than relying on large language models.

This project emphasizes:

• Practical application of NLP pipelines
• Classical ML over black-box APIs
• Explainability and transparency in language understanding

---

👥 Team Members

• Gautam N Chipkar
• Ameena Naik
• Asma Inamdar
• Jasmine Mulla

🚀 Overview

QueryGenie is a Natural Language Interface to Database (NLIDB) that enables users to query structured databases using plain English or voice.

Unlike LLM-based systems, it is built entirely using:

• 🐍 Python-based NLP pipeline
• 🧠 Classical Machine Learning (scikit-learn)
• 🧩 Rule-based linguistic processing

This makes the system:

• 🔒 Fully offline
• 💸 Zero-cost
• ⚡ Lightweight and fast
• 🛡️ Privacy-preserving

---

✨ Features

🧠 Natural Language → SQL

"Show students who scored more than 80"

SELECT * FROM STUDENT WHERE MARKS > 80;

---

🔐 100% Offline & Private

• No APIs, no cloud
• Runs locally
• Suitable for secure environments

---

🎯 NLP-Driven Understanding

• Intent Classification
• Entity Extraction
• Semantic Similarity Matching

---

📊 Explainable UI (Streamlit)

• Query results visualization

• Debug panel with:

  • Intent + confidence
  • Extracted entities
  • Generated SQL

---

🎤 Voice Input

• Speech-to-text query support

---

🧠 NLP Modules Covered

This project demonstrates key NLP concepts typically covered in a semester course:

1. Text Preprocessing

• Tokenization
• Lowercasing
• Stopword handling (implicit via TF-IDF)

2. Feature Extraction

• TF-IDF Vectorization
• Converts text into numerical feature space

3. Intent Classification

• Logistic Regression (Supervised Learning)
• Maps user queries to predefined intents

4. Semantic Similarity

• Cosine Similarity
• Handles ambiguous or unseen queries

5. Entity Extraction (Slot Filling)

• Regex-based pattern matching

• Extracts:

  • Numerical values (e.g., 80)
  • Conditions (>, <, =)
  • Limits (Top N queries)

6. Template-Based Language Understanding

• Maps structured intent + entities → SQL templates

7. Natural Language Generation (NLG)

• Converts SQL results into readable responses

---

🧩 System Architecture

flowchart LR
    A[User Input] --> B(TF-IDF Vectorization)
    B --> C(Intent Classifier)
    C --> D(Entity Extraction)
    D --> E(SQL Generator)
    E --> F(Database Execution)
    F --> G(Response Generator)
    G --> H(UI Output)

Loading

---

🛠️ Tech Stack

🧠 NLP Stack (Core Intelligence)

Component	Technology	Purpose
Text Vectorization	TF-IDF	Converts natural language into numerical features
Similarity Engine	Cosine Similarity	Handles ambiguous queries via semantic matching
Intent Classification	Logistic Regression (scikit-learn)	Predicts user intent from query
Entity Extraction	Regex (re)	Extracts conditions, values, and limits
NLP Pipeline	Tokenization, Feature Engineering, Slot Filling	End-to-end language understanding workflow
Response Generation	Rule-based NLG	Converts SQL output into readable responses

---

⚙️ System Stack (Execution Layer)

Component	Technology	Purpose
Language	Python 3	Core implementation language
Backend Logic	Modular Python Scripts	Handles pipeline orchestration
UI Framework	Streamlit	Interactive frontend + debug interface
Database	SQLite3	Local query execution engine
Data Handling	Pandas	Data processing and formatting
Voice Input	SpeechRecognition	Converts speech to text
Security	Template-based SQL	Prevents SQL injection

---

📂 Project Structure

QueryGenie/
│
├── app.py
├── intent_classifier.py
├── entity_extractor.py
├── sql_generator.py
├── response_generator.py
├── speech_handler.py
├── sql.py
├── student.db
├── requirements.txt
└── README.md

---

⚙️ Setup & Installation

git clone https://github.com/gee-46/querygenie.git
cd querygenie
python -m venv venv

Activate:

venv\Scripts\activate   # Windows
source venv/bin/activate # macOS/Linux

pip install -r requirements.txt
python sql.py
streamlit run app.py

---

💡 Example Queries

• "Show all students"
• "How many students are there?"
• "Top 3 performers"
• "Average marks"
• "Students scoring above 80"

---

🎯 Use Cases

• 🎓 Academic NLP demonstrations
• 📊 Database querying without SQL knowledge
• 🔐 Offline enterprise tools
• 🧠 Learning end-to-end NLP pipelines

---

⚠️ Limitations

• Single-table schema
• Limited intent set
• No advanced NER (yet)

---

🚀 Future Improvements

• spaCy-based Named Entity Recognition
• Multi-table JOIN support
• Offline speech models (Whisper/Vosk)
• Data visualization

---

👨‍💻 Author

Gautam N Chipkar GitHub: https://github.com/gee-46

---

⭐ Support

• Star ⭐
• Fork 🍴
• Build 🚀

---

📜 License

MIT License

---

💎 Core Idea

This project proves that powerful NLP systems can be built using Python and classical techniques — without relying on expensive APIs or large models.

Explainable. Offline. Academic. Practical.