IMPLEMENTASI CHATBOT PADA PENCARIAN SKRIPSI DI PERPUSTAKAAN TERINTEGRASI DENGAN WHATSAPP MENGGUNAKAN METODE LLM
Muhammad Adryan Hasbi; Rahmi Imanda;
No. Panggil: eSTI-2103015109
ABSTRAK
IMPLEMENTASI CHATBOT PADA PENCARIAN SKRIPSI DI PERPUSTAKAAN TERINTEGRASI DENGAN WHATSAPP MENGGUNAKAN METODE LLM
Muhammad Adryan Hasbi
Kesulitan mahasiswa dalam menemukan referensi skripsi yang relevan sering menjadi hambatan
dalam penyusunan tugas akhir. Penelitian ini mengembangkan chatbot pencarian skripsi berbasis
Large Language Model (LLM) yang terintegrasi dengan WhatsApp sebagai solusi interaktif. Data
awal berjumlah 1.018 skripsi dari perpustakaan FTII UHAMKA, kemudian melalui proses cleansing
dan validasi diperoleh 795 data yang digunakan. Sistem dibangun dengan framework LangChain
melalui modul Setting Table Schema, Semantic Search, Rank Result, dan Natural Language
Interface for Databases. Evaluasi dilakukan dengan pengujian akurasi dan User Experience
Questionnaire (UEQ). Hasilnya, chatbot mampu menjawab delapan skenario pertanyaan berbasis
bahasa alami dengan akurasi 100%. Pada UEQ yang melibatkan 26 responden, skor tertinggi
diperoleh pada skala Kejelasan (2,08) dan Ketepatan (2,00), sedangkan skala lain seperti Efisiensi
(1,92), Daya Tarik (1,88), Stimulasi (1,88), dan Kebaruan (1,88) juga menunjukkan nilai positif dan
masuk kategori “Excellent”. Temuan ini membuktikan bahwa chatbot tidak hanya akurat secara
teknis, tetapi juga unggul dari sisi pengalaman pengguna, serta berpotensi menjadi inovasi layanan
pencarian akademik berbasis percakapan di perguruan tinggi.
Kata kunci: Chatbot, LLM, LangChain, Pencarian Skripsi, WhatsApp.
Implementation of a Chatbot for Thesis Retrieval in a Library System Integrated with WhatsApp Using the LLM Method
Muhammad Adryan Hasbi
Difficulties in finding relevant thesis references often become an obstacle for students in completing
their final projects. This research develops a thesis retrieval chatbot based on a Large Language
Model (LLM) integrated with WhatsApp as an interactive solution. The initial dataset consisted of
1,018 theses from the FTII UHAMKA library, which, after cleansing and validation, resulted in 795
valid records used for system development. The system was built using the LangChain framework
through the modules of Setting Table Schema, Semantic Search, Rank Result, and Natural Language
Interface for Databases. Evaluation was carried out using accuracy testing and the User Experience
Questionnaire (UEQ). The results showed that the chatbot successfully answered eight natural
language query scenarios with 100% accuracy. In the UEQ evaluation involving 26 respondents,
the highest scores were obtained in the Clarity scale (2.08) and Accuracy scale (2.00), while other
scales such as Efficiency (1.92), Attractiveness (1.88), Stimulation (1.88), and Novelty (1.88) also
achieved positive scores, all categorized as “Excellent” according to the international benchmark.
These findings demonstrate that the chatbot is not only technically accurate but also excels in user
experience, and it has the potential to become an innovative conversational academic search service
in higher education institutions.
Keywords: Chatbot, LLM, LangChain, Thesos Search, WhatsApp
