Sistem Tanya Jawab Resep Nusantara Berbasis Qwen dan Retrieval-Augmented Generation
DOI:
https://doi.org/10.23960/komputasi.v14i2.373Keywords:
large language model, Qwen, retrieval-augmented generation, resep masakan Nusantara, tanya jawabAbstract
Pengembangan sistem question answering pada domain resep masakan menghadapi tantangan dalam menghasilkan jawaban yang koheren secara bahasa sekaligus konsisten terhadap bahan, langkah memasak, dan konteks prosedural dokumen sumber. Large Language Model (LLM) seperti Qwen memiliki kemampuan memahami dan menghasilkan teks, namun berpotensi menghasilkan hallucination ketika jawaban tidak didukung informasi relevan. Penelitian ini mengembangkan sistem tanya jawab resep masakan Nusantara berbasis Qwen dan Retrieval-Augmented Generation (RAG) untuk meningkatkan relevansi dan groundedness jawaban terhadap knowledge base resep. Dataset yang digunakan berasal dari Indonesian Food Recipes Dataset, diproses melalui tahap preprocessing, pembentukan data question-answer, fine-tuning model, dan pembangunan sistem retrieval berbasis vector database. Evaluasi dilakukan menggunakan metrik ROUGE, Recall@K, Grounded ROUGE, Grounded Ingredient Coverage, Answered Rate, dan No-answer Rate. Hasil evaluasi menunjukkan bahwa integrasi RAG meningkatkan kualitas dan keterikatan jawaban dibandingkan model tanpa retrieval, dengan performa terbaik diperoleh oleh skenario RAG + Fine-tuned Qwen yang mencapai ROUGE-L sebesar 0.413, Grounded ROUGE-L sebesar 0.926, dan Grounded Ingredient Coverage sebesar 0.967. Mekanisme tambahan yang dikonfigurasi pada sistem, yaitu no-answer filtering dan fallback retrieval, juga membantu menjaga konsistensi jawaban ketika konteks retrieval yang diperoleh kurang relevan. Penelitian ini diharapkan dapat menjadi langkah awal dalam pengembangan sistem question answering domain-spesifik berbasis LLM yang lebih konsisten dan grounded, khususnya pada pengetahuan prosedural seperti resep masakan Nusantara.
Downloads
References
Y. Gao et al., “Retrieval-augmented generation for large language models: A survey,” arXiv preprint arXiv:2312.10997, vol. 2, no. 1, 2023.
Y. Huang and J. X. Huang, “A survey on retrieval-augmented text generation for large language models,” ACM Comput. Surv., 2024.
A. Frummet and D. Elsweiler, “Qooka: A cooking question answering dataset,” in Proceedings of the 2024 Conference on Human Information Interaction and Retrieval, 2024, pp. 406–410.
A. Frummet, D. Elsweiler, and U. Kruschwitz, “Cooking with context: Leveraging context for procedural question answering,” Inf. Process. Manag., vol. 62, no. 6, p. 104212, 2025.
L. Huang et al., “A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions,” ACM Trans. Inf. Syst., vol. 43, no. 2, pp. 1–55, 2025.
M. S. Tamber et al., “Benchmarking llm faithfulness in rag with evolving leaderboards,” in Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track, 2025, pp. 799–811.
T. Dettmers, A. Pagnoni, A. Holtzman, and L. Zettlemoyer, “Qlora: Efficient finetuning of quantized llms,” Adv. Neural Inf. Process. Syst., vol. 36, pp. 10088–10115, 2023.
J. J. Pan, J. Wang, and G. Li, “Survey of vector database management systems,” arXiv preprint arXiv:2310.14021, 2023.
E. Burgin, S. Dutta, and M. Wang, “QuARK: LLM-Based Domain-Specific Question Answering using Retrieval Augmented Generation and Knowledge Graphs,” in Proceedings of the 15th International Conference on Recent Advances in Natural Language Processing-Natural Language Processing in the Generative AI Era, 2025, pp. 210–217.
R. Ahadi, N. S. Harahap, M. Fikry, and F. Kurnia, “Retrieval-Augmented Generation in a Web-Based Question Answering System for Fiqh Books,” Journal of Artificial Intelligence and Software Engineering (JAISE), vol. 5, no. 2, pp. 626–635, 2025, doi: 10.30811/jaise.v5i2.7005.
A. Stolfo, “Groundedness in retrieval-augmented long-form generation: An empirical study,” in Findings of the Association for Computational Linguistics: NAACL 2024, 2024, pp. 1537–1552.
I. T. Sorodoc, L. F. R. Ribeiro, R. Blloshmi, C. Davis, and A. de Gispert, “Garage: A benchmark with grounding annotations for rag evaluation,” in Findings of the Association for Computational Linguistics: ACL 2025, 2025, pp. 17030–17049.
C. Niu et al., “Ragtruth: A hallucination corpus for developing trustworthy retrieval-augmented language models,” in Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2024, pp. 10862–10878.
H. Joren, J. Zhang, C.-S. Ferng, D.-C. Juan, A. Taly, and C. Rashtchian, “Sufficient context: A new lens on retrieval augmented generation systems,” in International Conference on Learning Representations, 2025, pp. 20310–20334.
B. Hui et al., “Qwen2.5-coder technical report,” arXiv preprint arXiv:2409.12186, 2024.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Jurnal Komputasi
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
