Pengembangan dan Optimasi Intent Matching pada Virtual Assistant Mona dalam Platform Netmonk Menggunakan Model Embedding

Briliansyah, Tamam Fajar (2026) Pengembangan dan Optimasi Intent Matching pada Virtual Assistant Mona dalam Platform Netmonk Menggunakan Model Embedding. Project Report. [s.n.], [s.l.]. (Unpublished)

Text 5025231142-Project_Report.pdf - Accepted Version Restricted to Repository staff only Download (2MB) | Request a copy

Abstract

Virtual assistant merupakan teknologi yang banyak digunakan untuk meningkatkan kualitas layanan pada platform digital. Netmonk sebagai penyedia aplikasi monitoring jaringan di Indonesia menghadirkan virtual assistant bernama MONA untuk membantu pengguna dalam mengelola infrastruktur dan perangkat jaringan secara interaktif. Namun, sistem chatbot sebelumnya masih memiliki keterbatasan dalam memahami maksud pengguna karena menggunakan metode intent matching berbasis TF-IDF dan Support Vector Machine (SVM). Oleh karena itu, pada kegiatan Kerja Praktik ini dilakukan pengembangan dan optimasi sistem intent matching menggunakan pendekatan embedding berbasis model BAAI/bge-small-en-v1.5 yang diimplementasikan melalui FastEmbed dan ONNX Runtime.
Selain itu, dilakukan pengembangan fitur knowledge base otomatis yang memungkinkan sistem mengambil data dari help center Netmonk melalui proses scraping maupun upload dokumen berbasis Markdown, sehingga mempermudah pembaruan informasi. Pengembangan juga mencakup pemisahan layanan antara chatbot builder dan chatbot engine. Untuk meningkatkan performa, diterapkan mekanisme caching pada proses intent matching. Hasil pengembangan menunjukkan bahwa pendekatan embedding mampu meningkatkan akurasi pemahaman intent serta menghasilkan respons yang lebih relevan, sekaligus meningkatkan efisiensi dan performa sistem chatbot secara keseluruhan.
====================================================================================================================================
A virtual assistant is a technology widely used to improve service quality on digital platforms. Netmonk, as a network monitoring application provider in Indonesia, introduces a virtual assistant called MONA to assist users in managing infrastructure and network devices interactively. However, the previous chatbot system still had limitations in understanding user intent, as it relied on TF-IDF and Support Vector Machine (SVM)-based intent matching methods. Therefore, in this internship project, the intent matching system was developed and optimized using an embedding-based approach with the BAAI/bge-small-en-v1.5 model, implemented through FastEmbed and ONNX Runtime. In addition, an automatic knowledge base feature was developed, enabling the system to retrieve data from the Netmonk help center through scraping processes as well as Markdown-based document uploads, thereby simplifying information updates. The development also included separating the services between the chatbot builder and the chatbot engine. To further enhance performance, a caching mechanism was implemented in the intent matching process. The results show that the embedding-based approach improves intent understanding accuracy and produces more relevant responses, while also increasing the overall efficiency and performance of the chatbot system.

Item Type:	Monograph (Project Report)
Uncontrolled Keywords:	Chatbot, Intent Matching, Embedding, Virtual Assistant, Netmonk, Knowledge Base
Subjects:	T Technology > T Technology (General)
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Tamam Fajar Briliansyah
Date Deposited:	21 Apr 2026 08:16
Last Modified:	21 Apr 2026 08:16
URI:	http://repository.its.ac.id/id/eprint/132841

Actions (login required)

View Item