Key points are not available for this paper at this time.
Synthesizing music using Long Short-Term Memory (LSTM) Networks is a widely studied field of research. How the musical notes are abstracted and represented is one of the biggest problems of this process. In this work, a text-based solution, which represents several attributes of musical notes, is proposed as a solution to the mentioned problem. In this notation, numeric representations of the duration and the characteristics of the musical notes in piano pieces are textually combined. The advantages and disadvantages of using the proposed notation are discussed. The proposed method is applied to the created LSTM model, obtained results are discussed. The melodies generated by the trained model are compared to human-made melodies via a survey and the results are shared.
Memis et al. (Wed,) studied this question.