Khalish, Adjie Patrian (2025) Analisis Performa Berbasis Arsitektur Large Language Model Untuk Extractive Summarization. Other thesis, Universitas Islam Riau.

Text 213510709.pdf - Submitted Version Restricted to Registered users only Download (1MB) | Request a copy

Abstract

This research analyzes the performance of Large Language Model (LLM) models, namely T5 and Llama 3.2, in the extractive summarization task for Indonesian text. The background of this research is based on the increasing volume of digital information that causes difficulties in understanding text effectively. Extractive summarization is a solution to filter out important information by extracting the core information without changing the language of the original text. In this study, the T5 and Llama 3.2 models were tested using 1000 data from the IndoSum dataset and evaluated with the ROUGE metric to measure the performance of the resulting summaries. Pre-processing was done to clean and adjust the data before the models were trained using fine-tuning. The results show that the Llama 3.2 model has a more stable convergence compared to T5 in the fine-tuning process, while both models show competitive performance in generating text summaries. This study provides insights into the application of LLM in automatic information extraction and opens up opportunities for further development in Indonesian text processing.

Actions (login required)

View Item