Open AccessTransfer Learning of Pre-trained Transformers for Covid-19 Hoax Detection in Indonesian Language
Author(s) -
Lya Hulliyyatus Suadaa,
Ibnu Santoso,
Amanda Tabitha Bulan Panjaitan
Publication year - 2021
Publication title -
indonesian journal of computing and cybernetics systems
Language(s) - English
Resource type - Journals
eISSN - 2460-7258
pISSN - 1978-1520
DOI - 10.22146/ijccs.66205
Subject(s) - hoax , computer science , transformer , the internet , transfer of learning , indonesian , artificial intelligence , covid-19 , natural language processing , speech recognition , world wide web , linguistics , engineering , electrical engineering , medicine , philosophy , alternative medicine , pathology , voltage , disease , infectious disease (medical specialty)
Nowadays, internet has become the most popular source of news. However, the validity of the online news articles is difficult to assess, whether it is a fact or a hoax. Hoaxes related to Covid-19 brought a problematic effect to human life. An accurate hoax detection system is important to filter abundant information on the internet. In this research, a Covid-19 hoax detection system was proposed by transfer learning of pre-trained transformer models. Fine-tuned original pre-trained BERT, multilingual pre-trained mBERT, and monolingual pre-trained IndoBERT were used to solve the classification task in the hoax detection system. Based on the experimental results, fine-tuned IndoBERT models trained on monolingual Indonesian corpus outperform fine-tuned original and multilingual BERT with uncased versions. However, the fine-tuned mBERT cased model trained on a larger corpus achieved the best performance.