在電子商務快速發展與資訊日益龐大複雜的時代下，消費者面對大量商品常感到選擇困難，進而影響線上銷售成效，為解決線上購物決策困難，本研究設計並比較兩類對話式推薦系統架構：一方面探討在相同 RAG（Retrieval-Augmented Generation）架構下，GPT-4o-mini與Claude 3.5 Sonnet兩種大型語言模型（LLM）之生成差異；另一方面以GPT-4o-mini為基礎，分析RAG與結合知識圖譜推理之GraphRAG架構的回應品質與效能表現。
本研究透過寢具商品資料與多情境對話模擬，系統性評估兩大維度：(1)模型差異：比較GPT-4o-mini與Claude 3.5 Sonnet在相同RAG架構下的生成表現，結果顯示Claude 3.5 Sonnet於語言豐富性與多輪推理（如潮濕環境需求推導）方面表現更佳，而GPT-4o-mini則在回應穩定性與延遲時間（平均快1.1秒）具有明顯優勢；(2)架構差異：以GPT-4o-mini為基礎，驗證GraphRAG相較於RAG在複合條件推薦（如「怕冷+過敏」）的準確性與語境貼合度均有明顯提升，且回應時間減少12.02%。此結果證實結合知識圖譜推理能有效強化語境理解與系統擴充性，驗證LLM結合語意檢索與知識圖譜推理於推薦系統中的應用價值。

In the era of rapidly growing e-commerce and increasingly complex information, consumers often face decision-making difficulties when confronted with many similar products, negatively affecting online sales performance. To address this issue, this study designs and compares two types of conversational recommender system architectures. The first part investigates the generative differences between two large language models (LLMs)—GPT-4o-mini and Claude 3.5 Sonnet—under the same Retrieval-Augmented Generation (RAG) framework. The second part, based on GPT-4o-mini, compares the performance and response quality of the original RAG architecture and GraphRAG, which integrates knowledge graph reasoning.
Using real bedding product data provided by an industry partner, multiple simulated dialogue scenarios were designed to evaluate semantic relevance, language fluency, logical reasoning, and system latency. Experimental results indicate that, under the same RAG framework, Claude 3.5 Sonnet achieved higher language naturalness scores, while GPT-4o-mini demonstrated greater stability and lower average latency.Furthermore, GraphRAG outperforms RAG in multi-turn and attribute-based recommendation tasks, demonstrating higher scalability. This study validates the potential of integrating LLMs with semantic retrieval and knowledge graph reasoning in conversational recommender systems.

Amaral Teixeira de Freitas, B., & de Alencar Lotufo, R. (2024). Retail-GPT: leveraging Retrieval Augmented Generation (RAG) for building E-commerce Chat Assistants. arXiv:2408.08925. Retrieved August 01, 2024, from https://ui.adsabs.harvard.edu/abs/2024arXiv240808925A
Bruch, S., Gai, S., & Ingber, A. (2023). An Analysis of Fusion Functions for Hybrid Retrieval. ACM Trans. Inf. Syst., 42(1), Article 20. https://doi.org/10.1145/3596512
Burke, R. (2002). Hybrid Recommender Systems: Survey and Experiments.
Chen, Q., Lin, J., Zhang, Y., Ding, M., Cen, Y., Yang, H., & Tang, J. (2019, November). Towards Knowledge-Based Recommender Dialog System. In (pp. 1803-1813). Association for Computational Linguistics. https://doi.org/10.18653/v1/D19-1189
Chernev, A., Böckenholt, U., & Goodman, J. (2015). Choice overload: A conceptual review and meta-analysis. Journal of Consumer Psychology, 25(2), 333-358. https://doi.org/https://doi.org/10.1016/j.jcps.2014.08.002
Christakopoulou, K., Radlinski, F., & Hofmann, K. (2016). Towards Conversational Recommender Systems Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco, California, USA. https://doi.org/10.1145/2939672.2939746
Friedman, L., Ahuja, S., Allen, D., Tan, T., Sidahmed, H., Long, C., Xie, J., Schubiner, G., Patel, A., Lara, H., Chu, B., Chen, Z., & Tiwari, M. (2023). Leveraging Large Language Models in Conversational Recommender Systems. https://doi.org/10.48550/arXiv.2305.07961
Gao, Y., Sheng, T., Xiang, Y., Xiong, Y., Wang, H., & Zhang, J. (2023). Chat-REC: Towards Interactive and Explainable LLMs-Augmented Recommender System. arXiv e-prints, arXiv:2303.14524. https://doi.org/10.48550/arXiv.2303.14524
Goldberg, D., Nichols, D., Oki, B. M., & Terry, D. (1992). Using collaborative filtering to weave an information tapestry. Commun. ACM, 35(12), 61–70. https://doi.org/10.1145/138859.138867
Han, H., Shomer, H., Wang, Y., Lei, Y., Guo, K., Hua, Z., Long, B., Liu, H., & Tang, J. (2025). RAG vs. GraphRAG: A Systematic Evaluation and Key Insights. https://doi.org/10.48550/arXiv.2502.11371
Helicone. (2025). GPT‑4o‑mini vs Claude 3.5 Sonnet Benchmark Report. https://www.helicone.ai/blog/gpt-4o-mini-vs-claude-3.5-sonnet
Iyengar, S., & Lepper, M. (2001). When Choice is Demotivating: Can One Desire Too Much of a Good Thing? Journal of personality and social psychology, 79, 995-1006. https://doi.org/10.1037/0022-3514.79.6.995
Jannach, D., & Bauer, C. (2020). Escaping the McNamara Fallacy: Towards more Impactful Recommender Systems Research. AI Magazine, 41(4), 79-95. https://doi.org/10.1609/aimag.v41i4.5312
Koren, Y., Bell, R., & Volinsky, C. (2009). Matrix Factorization Techniques for Recommender Systems. Computer, 42(8), 30-37. https://doi.org/10.1109/MC.2009.263
Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., Küttler, H., Lewis, M., Yih, W.-t., Rocktäschel, T., Riedel, S., & Kiela, D. (2020). Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. arXiv:2005.11401. Retrieved May 01, 2020, from https://ui.adsabs.harvard.edu/abs/2020arXiv200511401L
Li, R., Kahou, S., Schulz, H., Michalski, V., Charlin, L., & Pal, C. (2018). Towards Deep Conversational Recommendations. arXiv:1812.07617. Retrieved December 01, 2018, from https://ui.adsabs.harvard.edu/abs/2018arXiv181207617L
Linden, G., Smith, B., & York, J. (2003). Amazon.com recommendations: item-to-item collaborative filtering. IEEE Internet Computing, 7(1), 76-80. https://doi.org/10.1109/MIC.2003.1167344
Lyu, H., Jiang, S., Zeng, H., Xia, Y., Wang, Q., Zhang, S., Chen, R., Leung, C., Tang, J., & Luo, J. (2024, June). LLM-Rec: Personalized Recommendation via Prompting Large Language Models. In (pp. 583-612). Association for Computational Linguistics. https://doi.org/10.18653/v1/2024.findings-naacl.39
Opoku, D. O., Sheng, M., & Zhang, Y. (2025). DO-RAG: A Domain-Specific QA Framework Using Knowledge Graph-Enhanced Retrieval-Augmented Generation. https://doi.org/10.13140/RG.2.2.22924.83844
Pazzani, M. J., & Billsus, D. (2007). Content-Based Recommendation Systems.
Pugazhenthi, T., & Liang, H. (2022, 17-20 Dec. 2022). Improving Conversational Recommender Systems via Knowledge Graph-based Semantic Fusion with Historical Interaction Data. In 2022 IEEE International Conference on Big Data (Big Data) (pp. 4303-4312). https://doi.org/10.1109/BigData55660.2022.10020689
Rendle, S. (2010, 13-17 Dec. 2010). Factorization Machines. In 2010 IEEE International Conference on Data Mining (pp. 995-1000). https://doi.org/10.1109/ICDM.2010.127
Resnick, P., Iacovou, N., Suchak, M., Bergstrom, P., & Riedl, J. (1994). GroupLens: An Open Architecture for Collaborative Filtering of Netnews.
Sanmartin, D. (2024). KG-RAG: Bridging the Gap Between Knowledge and Creativity. https://doi.org/10.48550/arXiv.2405.12035
Schwartz, B. (2004). The Paradox of Choice. Why More is Less.
Sun, Y., & Zhang, Y. (2018). Conversational Recommender System The 41st International ACM SIGIR Conference on Research & Development in Information Retrieval, Ann Arbor, MI, USA. https://doi.org/10.1145/3209978.3210002
Vellum. (2024). LLM benchmarks: Overview, limits, and model comparison. https://www.vellum.ai/blog/llm-benchmarks-overview-limits-and-model-comparison
Weaviate. (2023). Unlocking the Power of Hybrid Search - A Deep Dive into Weaviate's Fusion Algorithms. https://weaviate.io/blog/hybrid-search-fusion-algorithms
Wu, L., Zheng, Z., Qiu, Z., Wang, H., Gu, H., Shen, T., Qin, C., Zhu, C., Zhu, H., Liu, Q., Xiong, H., & Chen, E. (2023). A Survey on Large Language Models for Recommendation. https://doi.org/10.48550/arXiv.2305.19860
Zhang, Y., Yu, W., Zhang, E., Chen, X., Hu, L., Jiang, P., & Gai, K. (2024). RecGPT: Generative Personalized Prompts for Sequential Recommendation via ChatGPT Training Paradigm. arXiv:2404.08675. Retrieved April 01, 2024, from https://ui.adsabs.harvard.edu/abs/2024arXiv240408675Z
Zhou, J., Wang, B., He, R., & Hou, Y. (2021, November). CRFR: Improving Conversational Recommender Systems via Flexible Fragments Reasoning on Knowledge Graphs. In (pp. 4324-4334). Association for Computational Linguistics. https://doi.org/10.18653/v1/2021.emnlp-main.355