Ringkasan Ekstraktif Otomatis pada Berita Berbahasa Indonesia Menggunakan Metode BERT

Franky Halim (Program Studi Teknik Informatika, Universitas Kristen Petra Surabaya)
Liliana Liliana (Program Studi Teknik Informatika, Universitas Kristen Petra Surabaya)
Kartika Gunadi (Program Studi Teknik Informatika, Universitas Kristen Petra Surabaya)

Jurnal Infra
Vol 10, No 1 (2022)

In this modern era, information has become an important part of everyday life. In getting information several things can be done where one of them is by reading. With the increasing amount of information available on the internet, it is difficult for humans to keep abreast of developments. Online news is also one of the sources of information on the internet with a very large number and various topics. Reading the whole information sometimes also takes a long time. Therefore, it is necessary to make a summary of the available online news to reduce reading time and obtain relevant information. In this research, a summary of the news will be made by selecting important sentences from the news text. The method used in this research is Bidirectional Encoder Representations from Transformers with the addition of a transformer encoder layer.Based on the results of the tests that have been carried out, the pre-trained indolem/indobert-base-uncased model can produce the best F1-Score 57.17 for ROUGE-1, 51.27 for ROUGE-2, and 55.20 for ROUGE-L using abstractive reference and 84.46 for ROUGE-1, 83.21 for ROUGE-2, and 83.40 for ROUGE-L using extractive reference.