ANALISIS KINERJA KEYBART DAN TRANSFORMER UNTUK KEYPHRASE GENERATION PADA DATASET ARTIKEL ILMIAH

FARIZIM, AHMAD SUGAMA and Yusliani, Novi and Rachmatullah, Muhammad Naufal (2025) ANALISIS KINERJA KEYBART DAN TRANSFORMER UNTUK KEYPHRASE GENERATION PADA DATASET ARTIKEL ILMIAH. Undergraduate thesis, Sriwijaya University.

[thumbnail of RAMA_55201_09021282126091_cover.jpg]
Preview
Image
RAMA_55201_09021282126091_cover.jpg - Cover Image
Available under License Creative Commons Public Domain Dedication.

Download (120kB) | Preview
[thumbnail of RAMA_55201_09021282126091.pdf] Text
RAMA_55201_09021282126091.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (3MB) | Request a copy
[thumbnail of RAMA_55201_09021282126091_TURNITIN.pdf] Text
RAMA_55201_09021282126091_TURNITIN.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (4MB) | Request a copy
[thumbnail of RAMA_55201_09021282126091_0008118205_0001129204_01_front_ref.pdf] Text
RAMA_55201_09021282126091_0008118205_0001129204_01_front_ref.pdf - Accepted Version
Available under License Creative Commons Public Domain Dedication.

Download (908kB)
[thumbnail of RAMA_55201_09021282126091_0008118205_0001129204_02.pdf] Text
RAMA_55201_09021282126091_0008118205_0001129204_02.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (1MB) | Request a copy
[thumbnail of RAMA_55201_09021282126091_0008118205_0001129204_03.pdf] Text
RAMA_55201_09021282126091_0008118205_0001129204_03.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (386kB) | Request a copy
[thumbnail of RAMA_55201_09021282126091_0008118205_0001129204_04.pdf] Text
RAMA_55201_09021282126091_0008118205_0001129204_04.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (497kB) | Request a copy
[thumbnail of RAMA_55201_09021282126091_0008118205_0001129204_05.pdf] Text
RAMA_55201_09021282126091_0008118205_0001129204_05.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (1MB) | Request a copy
[thumbnail of RAMA_55201_09021282126091_0008118205_0001129204_06.pdf] Text
RAMA_55201_09021282126091_0008118205_0001129204_06.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (128kB) | Request a copy
[thumbnail of RAMA_55201_09021282126091_0008118205_0001129204_07_ref.pdf] Text
RAMA_55201_09021282126091_0008118205_0001129204_07_ref.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (139kB) | Request a copy
[thumbnail of RAMA_55201_09021282126091_0008118205_0001129204_08_lamp.pdf] Text
RAMA_55201_09021282126091_0008118205_0001129204_08_lamp.pdf - Accepted Version
Restricted to Repository staff only
Available under License Creative Commons Public Domain Dedication.

Download (90kB) | Request a copy

Abstract

Keyphrase play an important role in applications such as information retrieval, text summarization, recommendation systems, and indexing of scientific documents. However, the automated process of generating key phrases still faces major challenges, especially when faced with complex English scientific articles rich in technical terms. Transformer-based generative approaches often produce phrases that are either irrelevant or too general. This study aims to evaluate the performance of KeyBART, a variant of BART customized for the task of keyphrase generation, to overcome these problems. Using 30,000 scientific articles as training data, KeyBART's performance is compared with the standard Transformer and evaluated using the F1-Score metric. Results show that KeyBART consistently generates more accurate key phrases, with an F1@M of 20.09% and F1@5 of 24.02%. These findings suggest that KeyBART is more effective in understanding scientific context and more reliable for automatically generating key phrases in scientific articles.

Item Type: Thesis (Undergraduate)
Uncontrolled Keywords: Keyphrase Generation, KeyBART, Transformer, Self-Attention, Scientific Article
Subjects: P Language and Literature > P Philology. Linguistics > P98-98.5 Computational linguistics. Natural language processing
Q Science > Q Science (General) > Q334-342 Computer science. Artificial intelligence. Algorithms. Robotics. Automation.
Q Science > QA Mathematics > QA75-76.95 Calculating machines > QA76 Computer software
Divisions: 09-Faculty of Computer Science > 55201-Informatics (S1)
Depositing User: Ahmad Sugama Farizim
Date Deposited: 14 Jul 2025 01:45
Last Modified: 14 Jul 2025 01:45
URI: http://repository.unsri.ac.id/id/eprint/177722

Actions (login required)

View Item View Item