<p>Sentiment analysis is an important research area and the traditional text-based sentiment classification often overlooks the nuanced context of specific aspects and emotions. In this work, we propose GRAGLLM, a hybrid Graph Retrieval-Augmented Generation framework that integrates a knowledge graph with Large Language Models (LLMs) to enable multi-aspect, emotion-aware sentiment analysis. We construct a Neo4j knowledge graph integrated with Large Language Models (LLMs) using prompt-based learning for Explainable AI, where nodes represent reviews and aspect categories, and edges capture the semantic relationships between them. Each review is annotated with fine-grained emotions instead of simple polarity, providing a richer sentiment context. This structured graph is combined with neural embeddings to support hybrid semantic search. GRAGLLM utilizes a hybrid semantic retrieval technique, to dynamically extract contextually pertinent information by merging Cypher-based graph querying with vector similarity assessments on semantic embeddings. Interpretable aspect-specific sentiment outputs are produced by using acquired contextual comprehension which is then structured as prompts for activating LLM. Performance metrics indicate that neural generation integrated with structured knowledge representation achieves accurate, interpretable and emotion sensitive sentiment analysis, which closely aligns with reference summaries and reveals insightful aspect level sentiment trends. The experimental evaluation demonstrates GRAGLLM outperforms baseline models with improvement of 6% ,5% and 4% for LSTM, GPT-based LLMs, and Standard RAG, respectively. This result highlights the effectiveness of GRAGLLM in accurately capturing sentiment and emotion with enhanced interpretability. Future work will explore scaling this approach to larger, multilingual datasets and enhancing real-time retrieval and explanation capabilities.</p>