Alih Bentuk Kalimat Non-Formal menjadi Kalimat Formal Menggunakan Pendekatan Machine Translation

Abstract

The purpose of this study is the development of a model by applying the Short Term Memory algorithm to transform informal sentences into formal sentences and determine the supporting and inhibiting factors of the accuracy value generated in the machine learning process. The method used is action research. Software development uses an incremental model through analysis, design, implementation, and testing. The agreed results are part of the results of the study which tests the standard words into formal sentences and determines the factors of the accuracy value generated in the study. The BLEU evaluation determines whether or not sentences are predicted by measuring the closeness of the formal sentences from the dataset. This research resulted in a web-based application. The conclusion obtained is a training model that is built to produce value and loss with a value of 0.7011 and 3.7401. Based on these results, training models are not good enough. This causes several factors that lead to training overfitting models.