Data Augmentation and Fine-tuning to Improve IndoBART Performance - Dalam bentuk pengganti sidang - Artikel Jurnal

MUHAMMAD DAFFA FERDIANSYAH

Informasi Dasar

345 kali
24.04.175
621.382
Karya Ilmiah - Skripsi (S1) - Reference

Recently, the growth and dissemination of information on the internet is becoming faster and more massive. The information is in the form of online text documents, articles, news, and reviews. This has encouraged a lot of research on automatic text summarization, especially abstractive summarization. One of the pre-trained models used in abstractive summarization is BART, which has been trained with Indonesian data under the name IndoBART. However, IndoBART still has shortcomings, namely the unsatisfactory ROUGE metric value. In this research, data augmentation and hyperparameter fine-tuning are proposed for IndoBART, which can hopefully improve the performance of IndoBART, especially on the ROUGE metric value. Data augmentation is performed by modifying the data, which aims to increase the variety of data so that the machine can better produce summaries. To further improve the performance, hyperparameter fine-tuning is also performed to adapt the model to the data. The results are quite satisfactory, data augmentation and hyperparameter fine-tuning can increase the ROUGE value and overcome some of the problems that occur in abstractive summarization.

Subjek

ARTIFICIAL INTELEGENCE
ELECTRICAL ENGINEERING,

Katalog

Data Augmentation and Fine-tuning to Improve IndoBART Performance - Dalam bentuk pengganti sidang - Artikel Jurnal
 
 
Inggris

Sirkulasi

Rp. 0
Rp. 0
Tidak

Pengarang

MUHAMMAD DAFFA FERDIANSYAH
Perorangan
Suyanto
 

Penerbit

Universitas Telkom, S1 Informatika
Bandung
2024

Koleksi

Kompetensi

  • CII3L3 - PEMBELAJARAN MESIN LANJUT
  • CII4G3 - PEMROSESAN BAHASA ALAMI

Download / Flippingbook

 

Ulasan

Belum ada ulasan yang diberikan
anda harus sign-in untuk memberikan ulasan ke katalog ini