Deep Learning for the Applications of Music Industry
For the music industry, one of the valuable applications is to generate the lyrics of popular music. However, the lyrics generators are not comparable to lyricists now, that is, it is difficult to generate creative results like human beings. If the AI agent can create original and artistic music, which means that the machines achieve high-level human intelligence, "creativity". Popular music with its lyrics is a medium of expression, mainly conveying emotions and thoughts. However, the machines cannot understand the human feelings included in the texts. Therefore, the generated lyrics more lay on imitation than creation. It means the current methods are difficult to break through the existing framework. The main bottleneck of the current lyrics generators is the lack of contextual harmony of the music melody, the ambiguity and the emotional inconsistency of the lyrics, and the lack of creativity. To achieve the applications, deep embedding methods can be used to extract the deep representations of musical features that according to the essential elements of popular music creation. Notably, the musical deep representations can be useful inputs for the generative models. For the context, GPT-2 model is powerful for text generation. The conditional GPT-2 model can be used to generate lyrics according to the given style. For suitable for singing, the structure and rhyme of lyrics can modify by the use of a syntactic parser and a rhyme modification module. With automatic and human evaluations, the experimental results show that the proposed method can generate lyrics with high structural consistency, rhyme consistency, and originality according to the given music style.
Jia-Wei Chang is an assistant professor in Department of Computer Science and Information Engineering at National Taichung University of Science and Technology. Since January 2019, he is a Young Professionals Chair of the Institution of Engineering and Technology (IET) - Taipei Network. Since 2017, he is a consultant of NEXCOM Industry 4.0 Innovation Center. During February to July 2018, he was an adjunct assistant professor in Department of Engineering Science at National Cheng Kung University. He was a data scientist and project manager at IoT BU, Nexcom during 2016-2017. He received the Ph.D. degree from Department of Engineering Science, National Cheng Kung University in 2017. His research interests include natural language processing, internet of things, artificial intelligence, data mining, and e-learning technologies.
Dr. Jia-Wei Chang
Department of Computer Science and Information Engineering, National Taichung University of Science and Technology, Taiwan