Pengembangan Sistem Tanya Jawab Peraturan Teknologi Informasi Di Indonesia Menggunakan Table Augmented Generation (TAG)

Aziz, Ilham Abdul (2025) Pengembangan Sistem Tanya Jawab Peraturan Teknologi Informasi Di Indonesia Menggunakan Table Augmented Generation (TAG). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026211105-Undergraduate_Theses.pdf Restricted to Repository staff only Download (7MB) | Request a copy

Abstract

Di era modern, hukum memainkan peran penting dalam mengatur berbagai aspek kehidupan masyarakat. Namun, akses terhadap informasi hukum yang akurat sering kali menjadi tantangan, terutama bagi masyarakat umum yang tidak memiliki latar belakang hukum. Penelitian ini bertujuan untuk mengembangkan sistem tanya jawab berbasis hukum dengan mengintegrasikan Large Language Model (LLM) dan basis data hukum terstruktur melalui metode Table Augmented Generation (TAG) untuk meningkatkan akses informasi hukum yang lebih mudah dan akurat. Sistem ini dirancang untuk menjawab pertanyaan hukum dengan memanfaatkan data terstruktur dari tabel regulasi hukum. Data yang digunakan meliputi undang-undang dan putusan pengadilan terkait teknologi informasi dan elektronik yang diambil dari website JDIH komdigi dan BPK. Peraturan yang digunakan meliputi 63 peraturan dan 2.423 pasal. Dalam penelitian ini, dua model LLM yang digunakan adalah Llama 3.1b Instruct dari Ollama dan Claude 3.5 Haiku. Metode TAG dirancang untuk menghubungkan pertanyaan pengguna dengan data hukum terstruktur melalui proses query synthesis, query execution, dan answer generation. Hasil penelitian menunjukkan bahwa sistem TAG meningkatkan kinerja retrieval dan kualitas generasi jawaban. Pada tahap retrieval, Claude mencapai F1-Score terbaik yaitu 0.2196. Untuk generasi teks, model Claude juga menunjukkan faithfulness 0.869 dan answer accuracy 0.830. Secara keseluruhan, sistem TAG meningkatkan Rouge-L dari 0.21 menjadi 0.40, answer relevancy dari 0.70 menjadi 0.77, dan answer accuracy dari 0.33 menjadi 0.59 dibandingkan dengan LLM baseline.
========================================================================================================================
In the modern era, law plays an important role in regulating various aspects of people's lives. However, access to accurate legal information is often a challenge, especially for the general public who do not have a legal background. This research aims to develop a law-based question and answer system by integrating the Large Language Model (LLM) and structured legal database through the Table Augmented Generation (TAG) method to improve easier and more accurate access to legal information. The system is designed to answer legal questions by utilizing structured data from legal regulation tables. The data used includes laws and court decisions related to information technology and electronics taken from the JDIH website of Komdigi and BPK. The regulations used include 63 regulations and 2,423 articles. In this research, the two LLM models used are Ollama's Llama 3.1b Instruct and Claude 3.5 Haiku. The TAG method is designed to link user queries with structured legal data through the processes of query synthesis, query execution, and answer generation. The results show that the TAG system improves retrieval performance and answer generation quality. In the retrieval stage, Claude achieved the best F1-Score of 0.2196. For text generation, the Claude model also showed faithfulness of 0.869 and answer accuracy of 0.830. Overall, the TAG system improved Rouge-L from 0.21 to 0.40, answer relevancy from 0.70 to 0.77, and answer accuracy from 0.33 to 0.59 compared to the LLM baseline.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Sistem Tanya Jawab, Table Augmented Generation, Hukum, Teknologi Informasi, Regulasi, Question Answering System, Table Augmented Generation, Law, Information Technology, Regulations.
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence T Technology > T Technology (General) > T57.5 Data Processing Z Bibliography. Library Science. Information Resources > ZA Information resources > ZA4450 Databases
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Ilham Abdul Aziz
Date Deposited:	29 Jul 2025 07:50
Last Modified:	29 Jul 2025 07:50
URI:	http://repository.its.ac.id/id/eprint/122653

Actions (login required)

View Item