Turning Reviews into Real Insights: Emotion-Aware Sentiment Analysis with GraphRAG and Neo4j
summary
GraphRAG builds a Neo4j knowledge graph from 41,000 Skytrax reviews using Sentence-BERT embeddings, Cypher queries and a RAG layer with LLMs (tested with GPT-family and DeepSea), turning chunks of text into aspect nodes (seat comfort, cabin staff, in-flight entertainment) linked to Ekman-based emotions. Retrieval precision/recall and MRR run mostly above 90%, and a shown insight flags in-flight entertainment = 58% anger. The pipeline runs as a periodic or live stack and yields explainable, queryable aspect‑level emotion signals.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 1:09Problem and airline domain
- 3:19Why GraphRAG with Neo4j
- 4:23GraphRAG approach explained
- 6:55Dataset and experimental setup
- 8:23Knowledge graph construction
- 9:06Evaluation and model comparison
- 9:49Insights and airline actions
- 10:58Future directions
- 11:41Closing remarks and Q&A
related talks
36mGraphRAG Explained 🔥 Building Smarter LLMs
40mRAG Pipeline Evaluation
39mEnhancing RAG with Neo4j Knowledge Graph
31mBuilding a Customer-Intelligence Brain: How GraphRAG Turns Data into Decisions
24mEmpowering Lifelong Learning: Transforming Data into Insights With GraphRAG
01h 40mKnowledge Graphs & GraphRAG: Techniques for Building Effective GenAI Applications
17mEnhancing Chat-Based Customer Recommendations With GraphRAG and Hybrid Search
53mPractical GraphRAG - Making LLMs smarter with Knowledge Graphs