Penerapan Natural Language Understanding pada Sistem Chatbot Informasi Palsu (Hoax) Menggunakan Dual Intent And Entity Transformer (Diet) Classifier dan Fuzzy String Matching

Abstract

Fake information (hoaxes) is information that does not really happen and has a negative impact that can distort the interpretation, perception, and assumption of the reader. The spread of false information (hoaxes) in Indonesia continues to increase tremendously. This condition can be said to be in an alarming condition. If this condition is not handled quickly and appropriately, it will certainly increase the high public misunderstanding of an event. Therefore, an approach is needed to explain the truth of information. This research aims to develop a question and answer chatbot system about false information (hoax) by applying the Dual Intent and Entity Transformer (DIET) Classifier and Fuzzy String Matching methods. This research uses question and answer data from the website https://www.mafindo.or.id/ and Mafindo's youtube channel. This research uses search engines that can find information from the websites http://turnbackhoax.id/, https://www.liputan6.com, https://cekfakta.tempo.co, and https://www.kompas.com/. Tests conducted using the Dual Intent and Entity Transformer (DIET) Classifier method approach resulted in an accuracy of 0.9.