eprintid: 138139 rev_number: 24 eprint_status: archive userid: 217769 dir: disk0/00/13/81/39 datestamp: 2025-09-08 09:51:06 lastmod: 2025-09-08 09:51:06 status_changed: 2025-09-08 09:51:06 type: thesis metadata_visibility: show creators_name: Nadhief Athallah Isya, - creators_name: Rasim, - creators_name: Ani Anisyah, - creators_nim: NIM2106413 creators_nim: NIDN0025077409 creators_nim: NIDN0011089304 creators_id: nadhiefathallahi@upi.edu creators_id: rasim@upi.edu creators_id: anianisyah@upi.edu contributors_type: http://www.loc.gov/loc.terms/relators/THS contributors_type: http://www.loc.gov/loc.terms/relators/THS contributors_name: Rasim, - contributors_name: Ani Anisyah, - contributors_nidn: NIDN0025077409 contributors_nidn: NIDN0011089304 contributors_id: rasim@upi.edu contributors_id: anianisyah@upi.edu title: IMPLEMENTASI MODEL IndoBERT PADA CHATBOT KESEHATAN GIGI DENGAN ALGORITMA MASKED LANGUAGE MODEL DAN NEXT SENTENCE PREDICTION ispublished: pub subjects: QA subjects: RK divisions: ILKOM full_text_status: restricted keywords: Asisten Kesehatan Virtual, Chatbot, Pengolahan Bahasa Alami, BERT Virtual Health Assistant, Chatbot, Natural Language Processing, BERT note: https://scholar.google.com/citations?user=BTgx9jMAAAAJ&hl=en ID SINTA Dosen Pembimbing: Rasim: 5990962 Ani Anisyah: 6786982 abstract: Tingginya biaya konsultasi medis dan kurangnya distribusi tenaga medis yang merata di Indonesia menjadi hambatan signifikan bagi masyarakat dalam mengakses layanan kesehatan. Penelitian ini bertujuan untuk merancang dan mengembangkan chatbot kesehatan gigi menggunakan algoritma Large Language Model (LLM), khususnya BERT (Bidirectional Encoder Representations from Transformers). Dengan metodologi pengembangan perangkat lunak Waterfall, penelitian mencakup tahap analisis kebutuhan, desain, implementasi, dan pengujian sistem. Kebutuhan sistem diidentifikasi melalui studi literatur, observasi, dan konsultasi pakar, menghasilkan spesifikasi modul seperti registrasi pengguna, antarmuka percakapan, basis pengetahuan FAQ gigi, dan fitur rekam medis dalam bentuk PDF. Dataset dikembangkan dari korpus teks medis berbahasa Indonesia dan dilatih menggunakan pendekatan Masked Language Model (MLM) dan Next Sentence Prediction (NSP). Sistem chatbot dirancang dalam platform web yang mengintegrasikan model NLP berbasis IndoBERT dengan modul intent classification untuk mendukung percakapan lanjutan. Hasil pengujian menunjukkan respons cepat (rata-rata 541 ms), akurat, dan relevan, dengan hasil black-box testing yang valid serta throughput stabil. Validasi pakar menggunakan kuesioner skala Likert 1–5 menghasilkan skor rata-rata 3,61 dari empat aspek utama: relevansi, kejelasan, kesesuaian medis, dan kelengkapan informasi. Chatbot ini dinyatakan layak sebagai sarana konsultasi awal dan edukasi kesehatan gigi secara daring. The high cost of medical consultations and the unequal distribution of healthcare professionals in Indonesia pose significant barriers for the public in accessing healthcare services. This study aims to design and develop a dental health chatbot using a Large Language Model (LLM) algorithm, specifically BERT (Bidirectional Encoder Representations from Transformers). Employing the Waterfall software development methodology, the research encompasses requirement analysis, system design, implementation, and testing stages. System requirements were identified through literature review, observation, and expert consultation, resulting in specifications such as user registration, conversational interface, dental FAQ knowledge base, and medical record export in PDF format. The dataset was developed from Indonesian medical text corpora and trained using the Masked Language Model (MLM) and Next Sentence Prediction (NSP) approaches. The chatbot system was deployed on a web platform integrating the IndoBERT-based NLP model with an intent classification module to support extended conversations. Testing results demonstrate fast response times (average 541 ms), accurate and relevant answers, valid black-box testing outcomes, and stable throughput. Expert validation using a Likert-scale questionnaire (1–5) yielded an average score of 3.61 across four main aspects: relevance, clarity, medical appropriateness, and completeness of information. The chatbot is considered feasible as an initial consultation tool and an online dental health education medium. date: 2025-08-26 date_type: published institution: Universitas Pendidikan Indonesia department: KODEPRODI55201#Ilmu Komputer_S1 thesis_type: other thesis_name: other official_url: https://repository.upi.edu/ related_url_url: https://perpustakaan.upi.edu/ related_url_type: org citation: Nadhief Athallah Isya, - and Rasim, - and Ani Anisyah, - (2025) IMPLEMENTASI MODEL IndoBERT PADA CHATBOT KESEHATAN GIGI DENGAN ALGORITMA MASKED LANGUAGE MODEL DAN NEXT SENTENCE PREDICTION. S1 thesis, Universitas Pendidikan Indonesia. document_url: http://repository.upi.edu/138139/1/S_KOM_2106413_Title.pdf document_url: http://repository.upi.edu/138139/2/S_KOM_2106413_Chapter1.pdf document_url: http://repository.upi.edu/138139/3/S_KOM_2106413_Chapter2.pdf document_url: http://repository.upi.edu/138139/4/S_KOM_2106413_Chapter3.pdf document_url: http://repository.upi.edu/138139/5/S_KOM_2106413_Chapter4.pdf document_url: http://repository.upi.edu/138139/6/S_KOM_2106413_Chapter5.pdf