Question Answering System in the Domain of Early Childhood Education in Bahasa Indonesia

Abstract

The purpose of this study is to create a media for teaching staff in early childhood education schools, which is one of the features of the education robot answering system. Speech is used as input and output data from the system. The question answering system is a conversation data in the domain of early childhood education that is collected. Preprocessing stages are performed in the dataset to produce data that can be processed by the system. The question answering system uses the RNN algorithm with the Seq2Seq model. The highest results of the training process are 89.5% accuracy, precision 99.02%, and 70.5% recall. The response generation test also obtained an accuracy of 75%. The results of testing the response to the Questions according to the dataset produce maximum value.