Music generation with deep learning techniques

Abstract

With the advancement of artificial intelligence techniques in recent years, the task of music generation has gained much attention. Music is a type of sequential data comprising distinctive structures and comes in many various forms, which makes for an interesting problem that can be tackled using many different approaches. Emotions cannot be removed from music as the art form naturally invokes emotions, from composers to listeners. Generating emotive music has been explored by various researchers, interested to produce human-like sounds that can influence the feelings of people. However, there has been few research done on allowing users to control the music generated automatically. There are various ways that users can input information and textual data is one of the ways for users to input information to guide the direction in which the music should be generated. In this work, we propose a method to combine sentiments of textual data from users to generate suitable emotional music. A user study was conducted to evaluate the generated music, demonstrating that they are able to effectively convey the emotions present in the textual input.