Skripsi

SIMPLIFIKASI TEKS MENGGUNAKAN MODEL BART-BASE

---

Accessibility to scientific information is often hampered by complex language structures and diction, making it difficult for the general public to understand. This study aims to develop an automatic text simplification system using the fine-tuning method on the Pre-Trained Language Model BART-base to convert complex texts into simpler ones without reducing their main meaning. Using the WikiLarge dataset, this study evaluates eight training scenarios with variations in batch size and learning rate hyperparameters. The test results show that the best configuration is achieved in the scenario with a batch size of 8 and a learning rate of 5e-5, which produces a SARI score of 42.15 and a decrease in reading difficulty level (FKGL) to 9.51. In-depth analysis reveals that the loss divergence phenomenon in the best model is an indicator of effective lexical adaptation, where the model successfully performs creative paraphrasing and sentence splitting to significantly improve readability compared to simply copying references.

---

Availability

Detail Information

Series Title

-

Call Number

T1931672026

Publisher

Indralaya : Prodi Teknik Informatika, Fakultas Ilmu Komputer Universitas Sriwijaya., 2026

Collation

xvi, 130 hlm.; ilus.; tab.; 29 cm.

Language

Indonesia

ISBN/ISSN

-

Classification

006.350 7

Content Type

Text

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Prodi Teknik Informatika
Model Bart-Base

Specific Detail Info

-

Statement of Responsibility

MI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment