Brazilian Lyrics-Based Music Genre Classification Using a BLSTM Network
Organize songs, albums, and artists in groups with shared similarity could be done with the help of genre labels. In this paper, we present a novel approach for automatic classifying musical genre in Brazilian music using only the song lyrics. This kind o
- PDF / 1,058,765 Bytes
- 10 Pages / 439.37 x 666.142 pts Page_size
- 32 Downloads / 211 Views
stract. Organize songs, albums, and artists in groups with shared similarity could be done with the help of genre labels. In this paper, we present a novel approach for automatic classifying musical genre in Brazilian music using only the song lyrics. This kind of classification remains a challenge in the field of Natural Language Processing. We construct a dataset of 138, 368 Brazilian song lyrics distributed in 14 genres. We apply SVM, Random Forest and a Bidirectional Long Short-Term Memory (BLSTM) network combined with different word embeddings techniques to address this classification task. Our experiments show that the BLSTM method outperforms the other models with an F1-score average of 0.48. Some genres like gospel, funk-carioca and sertanejo, which obtained 0.89, 0.70 and 0.69 of F1-score, respectively, can be defined as the most distinct and easy to classify in the Brazilian musical genres context.
Keywords: Music genre classification Neural networks
1
· Natural language processing ·
Introduction
Music is part of the day-to-day life of a huge number of people, and many works try to understand the best way to classify, recommend, and identify similarities between songs. Among the tasks that involve music classification, genre classification has been studied widely in recent years [12] since musical genres are the main top-level descriptors used by music dealers and librarians to organize their music collections [9]. Automatic music genre classification based only on the lyrics is considered a challenging task in the field of Natural Language Processing (NLP). Music genres remain a poorly defined concept, and boundaries between genres still remain fuzzy, which makes the automatic classification problem a nontrivial task [9]. Traditional approaches in text classification have applied algorithms such as Support Vector Machine (SVM) and Na¨ıve Bayes, combined with handcraft c Springer Nature Switzerland AG 2020 L. Rutkowski et al. (Eds.): ICAISC 2020, LNAI 12415, pp. 525–534, 2020. https://doi.org/10.1007/978-3-030-61401-0_49
526
R. de Ara´ ujo Lima et al.
features (POS and chunk tags) and word count-based representations, like bagof-words. More recently, the usage of Deep Learning methods such as Recurrent Neural Networks (RNNs) and Convolutional Neural Networks (CNNs) has produced great results in text classification tasks. Some works like [4,5,7] focus on classification of mood or sentiment of music based on its lyrics or audio content. Other works, like [9], and [10], on the other hand, try to automatically classify the music genre; and the work [1] tries to classify, besides the music genre, the best and the worst songs, and determine the approximate publication time of a song. In this work, we collected a set of about 130 thousand Brazilian songs distributed in 14 genres. We use a Bidirectional Long Short-Term Memory (BLSTM) network to make a lyrics-based music genre classification. We did not apply an elaborate set of handcraft textual features, instead, we represent the lyrics songs with a pre-trained word em
Data Loading...
