Bert Transformer Model for Indonesian Mood Music Dataset

Music is an inseparable part of human life. Music can affect a person's mood. One's soul can unite in music. Music influences a person to feel sad, happy and relaxed. In terms of music, in general Indonesia has two types of music. Regional and national music. National music consists of several genres. Some of them are Dangdut, Pop, Ballad, and Rock. The many genres and developments in the world of music open the opportunities to the field of computer science to be explored further regarding to the amount of data contained in music processing (songs). In this study, the authors used self-processed data sets. The songs were downloaded from Youtube channel then were converted and only required chorus. In practice, the author used the Bidirectional Encoder Representations from Transformers (BERT) model. In learning, this model machine is used to improve accuracy. The resulting accuracy of this method is 81% with a chorus of more than 1000 songs.