knowledge-graph-NLP-pipeline

An end-to-end system that extracts entities and relationships from unstructured text (PDF, Text, HTML) and constructs an interactive Knowledge Graph in Neo4j.

Features

1. Document Processing

Load and clean PDF, Text, and HTML files.

2. NLP Pipeline & Knowledge Graph

Automated graph construction using Named Entity Recognition (BERT) and Relationship Extraction.

3. Interactive Visualization

Explore the graph with zooming and panning capabilities.

• Reasoning: Basic graph analysis (Centrality) using NetworkX.
• Visualization: Interactive Streamlit dashboard with PyVis graph view.

Prerequisites

• Python 3.8+
• Neo4j Desktop (running locally)

Installation

1. Clone the repository:

   git clone <repository-url>
   cd AntiGravity

2. Set up Virtual Environment:

   python -m venv venv
   .\venv\Scripts\activate  # Windows
   # source venv/bin/activate  # Mac/Linux

3. Install Dependencies:

   pip install -r requirements.txt
   python -m spacy download en_core_web_sm

4. Configure Environment: Create a .env file with your Neo4j credentials:

   NEO4J_URI=bolt://localhost:7687
   NEO4J_USER=neo4j
   NEO4J_PASSWORD=antigravity

Usage

Run the Streamlit application:

streamlit run src/app/main.py

Upload a document, click "Build Knowledge Graph", and explore the results.

Project Structure

• src/document_processing: Loaders and cleaners.
• src/nlp: NER and Relation Extraction models.
• src/graph: Neo4j interaction logic.
• src/reasoning: Graph analysis algorithms.
• src/app: Streamlit dashboard.

---