Brazilian Lyrics-Based Music Genre Classification Using a BLSTM Network
Authors:
Raul de Araújo Lima,
Rômulo César Costa de Sousa,
Simone Diniz Junqueira Barbosa,
Hélio Cortês Vieira Lopes
Abstract:
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 distrib…
▽ More
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.
△ Less
Submitted 6 March, 2020;
originally announced March 2020.
VisMaker: a Question-Oriented Visualization Recommender System for Data Exploration
Authors:
Raul de Araújo Lima,
Simone Diniz Junqueira Barbosa
Abstract:
The increasingly rapid growth of data production and the consequent need to explore data to obtain answers to the most varied questions have promoted the development of tools to facilitate the manipulation and construction of data visualizations. However, building useful data visualizations is not a trivial task: it may involve a large number of subtle decisions that require experience from their…
▽ More
The increasingly rapid growth of data production and the consequent need to explore data to obtain answers to the most varied questions have promoted the development of tools to facilitate the manipulation and construction of data visualizations. However, building useful data visualizations is not a trivial task: it may involve a large number of subtle decisions that require experience from their designer. In this paper, we present VisMaker, a visualization recommender tool that uses a set of rules to present visualization recommendations organized and described through questions, in order to facilitate the understanding of the recommendations and assisting the visual exploration process. We carried out two studies comparing our tool with Voyager 2 and analyzed some aspects of the use of tools. We collected feedback from participants to identify the advantages and disadvantages of our recommendation approach. As a result, we gathered comments to help improve the development of tools in this domain.
△ Less
Submitted 14 February, 2020;
originally announced February 2020.
