Identifikasi Berita Hoax Berbahasa Indonesia Menggunakan Bidirectional Long Short Term Memory (Bi-LSTM)

Abstract

There are still few hoax news that can be identified because it requires special knowledge, while there are still relatively few people who have this ability. Currently, the identification system is still done manually, so if more and more information is circulated, it will certainly become more difficult and troublesome. Therefore, an automatic system is needed that is able to identify news that falls into the hoax or non-hoax category. This research aims to implement the Bidirectional Long Short Term Memory (Bi-LSTM) algorithm to automatically identify hoax news titles in Indonesian. The model uses word embedding to represent text into vectors. The results of this research show that a model with accuracy 91% and a system is capable of identifying whether news headlines are hoaxes or non-hoaxes.