ISLAM, Muhamad Anbiya Nur and Warsito, Budi and Nurhayati, Oky Dwi (2024) PENGEMBANGAN CHATBOT TANYA JAWAB OTOMATIS BERBASIS NATURAL LANGUAGE PROCESSING DENGAN PATTERN MATCHING DAN SEQUENCE-TO-SEQUENCE. Masters thesis, UNIVERSITAS DIPONEGORO.

	Text 1. Cover-1.pdf Download (105kB)
	Text 1. Cover.pdf Restricted to Repository staff only Download (6MB)
	Text 2. BAB_I.pdf Download (553kB)
	Text 3. BAB_II.pdf Download (5MB)
	Text 4. BAB_III.pdf Restricted to Repository staff only Download (1MB)
	Text 5. BAB_IV.pdf Restricted to Repository staff only Download (6MB)
	Text 6. BAB_V.pdf Restricted to Repository staff only Download (345kB)
	Text 7. Daftar_Pustaka.pdf Download (1MB)

Abstract

Universitas Negeri Semarang (UNNES) menghadapi tantangan peningkatan volume permohonan layanan di Unit Layanan Terpadu pasca pandemi COVID-19, yang berisiko menurunkan efisiensi serta kepuasan pengguna. Penelitian ini bertujuan menjawab keterbatasan kapasitas customer service (CS) dengan mengembangkan chatbot tanya jawab otomatis berbasis Natural Language Processing (NLP). Metode yang digunakan adalah kombinasi pattern matching berbasis Term Frequency-Inverse Document Frequency (TF-IDF) dan model sequence-to-sequence (seq2seq). Dataset berupa 4.000 query disederhanakan menjadi 200 Frequently Asked Questions (FAQ), kemudian dibersihkan dan diproses untuk menemukan metode terbaik. Hasil menunjukkan TF-IDF unggul dengan akurasi 78% dan waktu pemrosesan rata-rata 0,01 detik untuk 50 pertanyaan, sementara seq2seq digunakan bagi pertanyaan berstruktur lebih kompleks dengan rata-rata BLEU Score 87,12. Temuan tersebut menghadirkan sistem informasi yang lebih responsif dan akurat, terbukti dari kepuasan pengguna sebesar 79,18% dan penurunan beban kerja CS secara signifikan. Dampak positif dari penelitian ini terlihat pada efisiensi pelayanan publik di UNNES, yang semakin optimal dan dapat fokus pada penanganan kasus yang lebih kompleks.
Kata Kunci: Natural Language Processing, Chatbot, Pattern Matching, Sequence-to-Sequence, TF-IDF, Frequently Asked Questions

Universitas Negeri Semarang (UNNES) faces escalating service requests at its Integrated Service Unit following the COVID-19 pandemic, posing risks to both efficiency and user satisfaction. This study aims to address the limited capacity of customer service (CS) staff by developing an automated question-answering chatbot based on Natural Language Processing (NLP). The proposed method integrates pattern matching using Term Frequency-Inverse Document Frequency (TF-IDF) with a sequence-to-sequence (seq2seq) model. A dataset of 4,000 user queries was condensed into 200 Frequently Asked Questions (FAQ), which were subsequently cleaned and pre-processed to identify the optimal approach. Results indicate that TF-IDF excels with 78% accuracy and an average processing time of 0.01 seconds for 50 queries, while seq2seq is deployed for more complex questions, achieving a mean BLEU Score of 97.70. These findings yield a more responsive and accurate information system, as reflected by a user satisfaction rate of 79.18% and a substantial reduction in the workload of CS personnel. The positive impact of this research is demonstrated by increased efficiency in public services at UNNES, enabling staff to focus more effectively on complex issues and enhancing overall service quality.
Keywords: Natural Language Processing, Chatbot, Pattern Matching, Sequence-to-Sequence, TF-IDF, Frequently Asked Quest

Actions (login required)

View Item