A course by EPFL
The universe of music is constantly growing. Thousands new songs are published each day all around the world. Moreover, the world is becoming more connected and this helps to spread music popularity.
We want to understand how music evolved in the last roughly 60 years, as music is an important social factor. People express themselves through music, and lot of cultural events and customs are derived from it. This popularity could help to explain some movements influences and the creation of some music styles. We have used the dataset from MillionSongDB with 450.000 songs, coupled with data from websites such as The Echo Nest (now Spotify) and Musicbrainz.
Music is a form of art, and a cultural activity which has been practiced since prehistoric time. Nowadays it is everywhere. Music seems inseparable from our life-style, as even humming or hitting an object with your finger makes "music". The evolution of this field is thus an interesting subject to study. It is linked to cultural movement in history, such as the 80's with hippie movement but also architecture (Baroque) or religion (Jean-Sébastien Bach). This can help us to understand how music evolves with the time, how events and epochs influence it. Music has also become a big industry, with the apparition of the term "commercial music". More and more people are interested in music e.g. this video with content based on music, has reached more than 100 million views. In this study we will concentrate on music from the 20th century and beginning of the 21th century. We asked ourselves the following questions:
Before going into details, we will spend a few words on the background of this blog post: The analyzed dataset consisted of roughly 460.000 Songs. After we cleaned the dataset, the remaining 220.000 songs were analyzed. The most important features used in the analysis are the following : tempo, duration, loudness, hotness of Songs and genres and genre weight of artists. Unfortunately, interesting features like energy and danceability were not usable from the dataset ( zero for each song ). For other features like key and rhythm of songs it was not possible to draw meaningful conclusions.
Presented increases and decreases in this blog post might deviate from the values presented in the charts. The charts present the average features per year, while the change rates have been calculated using confident intervals to ensure correctness.
In this section, we will concentrate on the evolution of music in general, meaning we did differentiate genres yet. The three most influenced features by time are tempo, loudness and duration.
For the tempo one can notice a seven BPM increase from 1960 to 1980. After 1980 the tempo drops by seven BPMs around 1992. In recent years one can notice a BPM increase of two BPM
Scroll through time to display are more detailed analysis!Loudness is defined as follows : "That attribute of auditory sensation in terms of which sounds can be ordered on a scale extending from quiet to loud.", or more theoretically, it is the subjective perception of sound pressure.
One can notice a steep increase of loudness of 5 dB. The steepest point of the curve is in the 90s, during the Loudness War
Scroll through time to display are more detailed analysis!For the duration of songs one can notice a very steep increase in the 60s and 70s. During those years the song Duration increased by almost one minute. After the 70s the duration kept increasing until the 2000s, when the duration started to decrease.
Scroll through time to display are more detailed analysis!Starting from this section we will consider the evolution of music based on different genres. The chart shows the portion of songs released per year per genre. The Genre matching was done ourselves to reduce several hundred genres to 14 genres.
We can clearly see the trending genres moving with the time. Hover over the lines in the chart to display descriptions for each genre.From this we can understand the popularity of some old time such as Jazz, Blues or Folk has greatly reduce since the beginning of the 20th century. This is due to the apparition of new genre based on this previous one. Also, cultural movement or apparition of some famous people increased the popularity of a genre with the example of Bob Marley for the Reggae or the rise of Metal/ Heavy-Metal in the late 70s. And finally, the technological advance has helped some genre such as Electronic to spread faster and take a place in this genre population.
In this section we will apply the gained knowledge of genres to the already evaluated features.
The evolution of genre loudness clarifies our observation of Loudness War shown in the previous plots. Indeed, there is a clear augmentation of loudness, with a higher change for metal, that might be due to the use of distortion and other procedure that increase the song force. We can however see 3 genres that are less affected which are Folk, Blues and Jazz. This should be due to the acoustic nature of these genre as it does not use as much electronical material as the other genres.
We have not found matching events or information for this regrouping, but there is some interesting fact to notice. Downtempo and Electronic has a linear increased, mostly due to their close relationship with technologic discovery. And, Metal and Reggae were the one with the lowest variation , it might be linked to the fact that it Metal appeared later than the other genre and used a high Loudness since the beginning and Reggae used maybe less this loudness increased.
So if we study the evolution of Tempo by genre, most of them the same trend as the average music, with some exceptions. Indeed Jazz, Folk and Blues music has the same shape of curve but shifted to lower BPM after the first maximum. This is due to the slow characteristic of these genre which are typically acoustic. It's indeed amazing how they almost perfectly match the Reggae has a trend to go a little higher in term of tempo, this is maybe due to the apparition of the Ska subgenre. The curve goes up to 128 instead of 125 and stay a little higher than the other genre. As we have seen previously, Metal is a real special genre with the tendency to never match the other genres. So, it's curve is like the Reggae one, but shifted even higher with a BPM that went to 132 and finish at almost 130
So here, for all genre we clearly see and increasing trend, with waves.
The analysis gives almost the same result as the previous one on Loudness.Our analysis has shown that independant of the genre, the tempo, loudness and duration of songs increased. We were able to explain the changes of features by e.g popularity changes of music genres or technological innovation. Furthermore, we were able to infer the genre popularity over time from the dataset. And finally, we identified noticeable differences of e.g. the song tempo evolution of different genres.