A general method to automatically generate music playlists

• Ano de defesa: 2019
• Autor(a) principal: Marcos Alves de Almeida
• Orientador(a): Renato Martins Assunção
• Banca de defesa: Olga Nikolaevna Goussevskaia, Flávio Vinícius Diniz de Figueiredo, Nazareno Ferreira de Andrade
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Minas Gerais
• Programa de Pós-Graduação: Programa de Pós-Graduação em Ciência da Computação
• Departamento: ICX - DEPARTAMENTO DE CIÊNCIA DA COMPUTAÇÃO
• País: Brasil
• Palavras-chave em Português: Playlist Generators; Music Embedding
• Link de acesso: http://hdl.handle.net/1843/31794

Music is one of the most used forms of entertainment, being consumed by people all over the world. Different from other types of entertainment such as movies and plays, music is consumed in playlists, that is, several tracks are grouped together before the users listen to them. Arranging the songs in a sequence is a time-consuming task, and may require specific knowledge from the playlist creator. The objective of this work is to propose a general method to automatically generate music playlists satisfying conflicting goals. First, we will analyze users' playlists in order to understand their characteristics and music genres. Next, we will propose methods to calculate the similarity between songs using acoustic characteristics and metadata. The proposed similarity functions will be used to embed the songs in a music space, where similar songs are close to each other. Then, we will propose a general method to automatically generate a random playlist of songs connecting two anchor songs defined by the user. Based on the general method, we will construct two algorithms to generate music playlists, named ROPE and STRAW, and apply them to the constructed music spaces. With the experiments carried out, we showed the proposed algorithms are able to generate random heterogeneous music playlists with smooth transitions between songs. Finally, an online prototype is developed to allow users to test the proposed method.