Rachma Adzima Maulida, . (2026) IMPLEMENTASI LARGE LANGUAGE MODEL DENGAN PENDEKATAN RETRIEVAL-AUGMENTED GENERATION PADA CHATBOT KONSULTASI DAN REKOMENDASI PRODUK HERBAL HNI-HPAI BERBASIS WEBSITE. Skripsi thesis, Universitas Pembangunan Nasional Veteran Jakarta.
![[img]](http://repository.upnvj.ac.id/style/images/fileicons/text.png) |
Text
ABSTRAK.pdf Download (209kB) |
|
Text
AWAL.pdf Download (931kB) |
|
Text
BAB 1.pdf Restricted to Repository UPNVJ Only Download (274kB) |
|
Text
BAB 2.pdf Restricted to Repository UPNVJ Only Download (745kB) |
|
Text
BAB 3.pdf Restricted to Repository UPNVJ Only Download (951kB) |
|
Text
BAB 4.pdf Restricted to Repository UPNVJ Only Download (1MB) |
|
Text
BAB 5.pdf Restricted to Repository UPNVJ Only Download (196kB) |
|
Text
DAFTAR PUSTAKA.pdf Download (132kB) |
|
Text
RIWAYAT HIDUP.pdf Restricted to Repository UPNVJ Only Download (90kB) |
|
Text
LAMPIRAN.pdf Restricted to Repository UPNVJ Only Download (6MB) |
|
Text
HASIL PLAGIARISME.pdf Restricted to Repository staff only Download (23MB) |
|
Text
ARTIKEL KI.pdf Restricted to Repository staff only Download (450kB) |
Abstract
The increasing use of herbal products in Indonesia has been accompanied by a growing public need for information that is rapid, relevant, and supported by clearly identified sources to assist in selecting products appropriate to specific health concerns. This study aims to design, implement, and evaluate HerbaAI, a web-based chatbot for consultation and recommendation of HNI herbal products that integrates the LLaMA 3.1 8B Instant model with a Retrieval-Augmented Generation (RAG) approach. The system was developed using the Rapid Application Development method, consisting of the Requirements Planning, User Design, Construction, and Cutover stages. The knowledge base was constructed from 306 documents on diseases and HNI products through preprocessing, chunking, embedding generation using intfloat/multilingual-e5-large-instruct, and storage in ChromaDB. The system was developed using Next.js, FastAPI, LangChain, SQLite, and the Groq API. Evaluation was conducted through black-box testing, ROUGE, and RAGAS using a golden dataset containing 20 questions. The RAG-based system achieved ROUGE-1, ROUGE-2, and ROUGE-L scores of 0.4998, 0.3533, and 0.4437, respectively, outperforming the non-RAG system. The RAGAS evaluation yielded scores of 0.8440 for faithfulness, 0.9518 for answer relevancy, 0.9211 for context precision, and 0.7504 for context recall. The results indicate that HerbaAI is capable of generating relevant and context-supported responses, although the coverage of retrieved context still requires improvement.
| Item Type: | Thesis (Skripsi) |
|---|---|
| Additional Information: | [No.Panggil: 1910511024] [Pembimbing: Zatin Niqotaini, S.Tr.Kom., M.Kom.] [Penguji 1: Musthofa Galih Pradana, M.Kom.] [Penguji 2: Anis Fitri Nur Masruriyah, S.Kom., M.Kom.] |
| Uncontrolled Keywords: | Chatbot, Herbal Medicine, LLaMA 3.1, Retrieval-Augmented Generation, RAGAS |
| Subjects: | Q Science > QA Mathematics > QA75 Electronic computers. Computer science Q Science > QA Mathematics > QA76 Computer software R Medicine > R Medicine (General) |
| Divisions: | Fakultas Ilmu Komputer > Program Studi Informatika (S1) |
| Depositing User: | RACHMA ADZIMA MAULIDA |
| Date Deposited: | 30 Jul 2026 07:32 |
| Last Modified: | 30 Jul 2026 07:32 |
| URI: | http://repository.upnvj.ac.id/id/eprint/51633 |
Actions (login required)
 |
View Item |