Penerapan Algoritma Neural Network Pada Chatbot Bahasa Jawa Tingkat Tutur Krama Alus

Abstract

Krama Alus are a level of speech in Javanese. In speaking or engaging in conversation with Krama Alus in everyday life, youths in particular school students are still unfamiliar and sometimes have difficulty answering with Krama Alus language. The study led to a web-based app called Krama Alus Chatbot using a Neural Network algorithm, which can respond to input from users using the Krama Alus Javanese. The purpose of the study is to know the implementation of the Neural Network algorithm in the making of chatbot, as well as to train the chatbot to respond to Krama Alus. The function of the algorithm in this study is to classify words put in by the user, to match sentences in established patterns, and to predict with answers matching the pattern. The patterns in the chatbot are shaped by conducting an interview with a Javanese teacher, and they are stored in a .JSON file. The Neural Network training process gets a reasonably high grade of accuracy with an average accuracy of 0.9. Chatbot can respond to input that matches the pattern quite well. Testing done using usability test can get a good predicate with an average of 72.8%.