Measuring emotions to classify songs: The impact of the COVID-19 pandemic on music streaming data

Abstract

Data from music streaming has gained increasing attention since it allows
studying music preferences across diverse cultures and different periods of time.
Indeed, the study of “music and emotion” is crucial for understanding the psychological
relationship between human sentiments and music. The temporal study
of musical emotions provides beneficial insights into the analysis of the mood of
listeners during periods of particular relevance and stress (e.g., the COVID-19
pandemic). This study performs music streaming data analysis to retrieve the
musical emotions of the top Italian streamed songs during the pandemic. To this
end, we propose two new indices for measuring anger and joy in songs. We suggest
a procedure for clustering music streaming data: the DISTATIS procedure
and Partitioning Around Medoids (PAM) clustering algorithm are sequentially
applied to identify intervals of time sharing similar sentiments. Finally, we employ
the proposed procedure to investigate the relationship between the evolution
of the pandemic spread and sentiments extracted from songs.

The results show that music streaming data analysis allow identifying five
clusters of time intervals sharing similar sentiments,