A CASE STUDY IN MUSIC STYLE'S SIMILARITY MEASUREMENT USING DEEP LEARNING

Ian Carvalho

Undergraduate dissertation thesis submited to Universidade de São Paulo as a requirement for obtaining the Computer Science Bachelor Degree

Abstract

This work tries to establish a similarity measure for music genres in Portuguese utilizing a deep neural network. It shows the necessary steps in order to gather and process the lyrics with a crawler and them proceeds to use this dataset in order to train two neural networks utilizing Long Short-Term Memory and Hierarchical Temporal Memory architectures. From the confusion matrices generated by those classifiers, this works then calculates a similarity matrix that indicate the likeness between two genres.

• Poster
• Dissertation
• Old website