Building Knowledge Graphs - LLM Enhanced Approach for Educational Video Recommendations

This tutorial demonstrates how to build an intelligent knowledge graph for educational video recommendations by combining Large Language Models (LLMs) with semantic embeddings. Learn to extract key concepts, difficulty levels, prerequisites, and learning outcomes from educational videos while establishing meaningful relationships between them to enable semantic search and personalized learning path generation. The hybrid approach leverages both LLMs' structured information extraction capabilities and embeddings' semantic similarity detection to deliver contextually relevant video recommendations even for complex queries. Follow along as the system processes video metadata through LLMs, generates vector embeddings, constructs a knowledge graph with meaningful relationships, and implements multiple query methods including semantic search, LLM-based understanding, and pattern matching fallback. This approach offers significant improvements over traditional NLP methods through deeper contextual understanding, concept expansion, and robust query handling. Access the complete code on GitHub and download the input CSV file from the provided links to implement this powerful recommendation system yourself.