The starting point of this post is a simple question: can we use R to analyze punk bands ? And, as a result: what can we learn from applying data analytics methods to punk music ?

Whether we like it or not “punk rock is arguably the most important subgenre of music to come out of the ‘70s” and consequently still an integral part of our mainstream contemporary culture. After years of being declared too outrageous to be accepted, its legacy is so astonishingly extensive that it deserves careful consideration and serious attention. Since decades, many music critiques, fine arts experts, social and political scientists or historians of pop culture have devoted time and energy to study the punk scene, its cultural production and legacy, the attitude of the punk generation, its tangle of ideologies, the ways it was perceived and received. Facts and figures, however, are still missing, perhaps because there apparently is nothing more distant from data analytics than punk music. So, is data analytics of punk rock possible ? Would it make any sense ? My answer is a loud and bold yes –yes, statistics on punk rock matters.

Although the punk scene cannot be condensed into a single band, the Ramones are still considered by many as the first “pure punk band” and, perhaps –and more importantly–, one of the most influential. This does not imply that other punk rock bands (Clash, Dead Kennedys, The Stooges, Misfits, Sex Pistols, Social Distorsion, Patti Smith Group, etc) are less noteworthy or not as good. Yet, since I need to start somewhere, I decided that my first attempt would focus on the Ramones –which I paradoxically like a lot despite being more of a baroque and classical music person.

What did the Ramones do ?

From 1976 to 1994, the Ramones released 14 studio albums. In their original USA release, the albums comprised 176 different songs in total that were quite short (median: 2M 32S) and mostly written in a Major key (only 2 songs are in a minor key: Em).

Musical purists always reproached the Ramones for knowing a couple of chords only and making an excessive use of them. Data show that the band knew at least… 11 different chords (out of too-many-to-bother-counting possibilities) although 80% of their songs were built on no more than 6. And there is no evidence of a sophistication of the Ramones’ compositions over time.

Just as the number of different chords in a Ramones’ song is independent from the song writer/s –t.test of number of different chords ~ writers don’t allow to exclude alternative hypothesis–, even with each band member having a very distinct personality, according to the biographers.

In terms of official charts ranking in the USA, the success of the Ramones fluctuated over their career. The first years of the band were definitely the most successful, from the creation of the band till the early 80’s. Then, from 1985 onwards, it looks like that the sales didn’t follow the strengthening of their reputation not only within but also outside the punk rock scene.

What did the Ramones say ?

Im my dataset, the Ramones’ lyrics come from azlyrics.com. I preferred this source over many other available sources since that website provides the lyrics without the verses repeats, which, in my opinion, would over-emphasise and, ultimately, biais the relevance of n-grams or topics. The dataset (a data frame) contains a lyrics variable, i.e. a character string of the track (without the verses repeats) including the < br> tags to mark the end of each line.

An example of the lyrics variable is like the following:

Hey ho, let s go < br>Hey ho, let s go < br>They re forming in a straight line < br>They re going through a tight wind < br>The kids are losing their minds < br>The Blitzkrieg Bop < br>They re piling in the back seat < br>They re generating steam heat < br>Pulsating to the back beat < br>The Blitzkrieg Bop. < br>Hey ho, let s go < br>Shoot em in the back now < br>What they want, I dont know < br>They re all reved up and ready to go

Tidying the text up (adopting the data principles recommended by Hadley Wickham) is the necessary first step of the lyrics mining exercise. For that, I follow the tidy text approach developed by Julia Silge & David Robinson.

First and foremost, it is worth noting that whatever the Ramones say, they say it in very few words ! Ramones songs are brief in time, but also short in lyrics (but not so much in vocabulary with 2,139 different unique words in total).

Whereas uniGrams are usually considered suitable for analysis after expurgation of stop words, in the Ramones lyrics the raw uniGrams show an interesting pattern. The 2 most frequent words in the 14 studio albums are i and you. One could provocatively argue that Tea for Two, a well-known 1925 song from Vincent Youmans and Irving Caesar, is a good representation of the Ramones musical universe that seems to be mainly centered on you and i, and i and you !

In the uniGrams table below, the columns of the cleaned uniGrams highlight that the top word in the Ramones lyrics is dont, expressing an atmosphere of clear negation. But there is also a fascinating tension pointing to the future that shows through words such as wanna, gonna and ll (will or shall). Rock and punk amongst the top 20 words definitely remind you what type of music you are listening to but also what subculture the band belongs to. In an all-men band, words such as baby, love, girl witness the significance of man-woman relationships in the Ramones songs. Perhaps it took statistical analysis of lyrics to take the risk of forming the hypothesis of the Ramones as a romantic band…

The identification of most frequent uniGrams per album is a further step into a more granular analysis:

In addition to identifying the most frequent single words, we could also highlight when they are used in the discography using a simple Token Distribution Analysis. Let’s limit this exercise to 5 words only from the list of the top 20: love, gonna, rock (or rocker), life and dont.

A quick visualisation of ‘raw’ nGrams (stop words not removed) confirms the feeling of a narrative universe mainly focused on i, you and negation (don’t).

What did the Ramones feel ?

As a (brief) final chapter of this post, I would like to run a very quick –and limited– sentiment analysis of the Ramones’ studio albums lyrics. Actually, rather than a sentiment analysis, this is nothing but scratching the surface of sentiment analysis. The bing sentiment lexicon was used here, but a similar analysis could be carried out using afinn or nrc lexicons (all available in the tidytext r package) or using all of them for a comparative approach.

Although the sentiment lexicon gives the word punk a negative value, there is little risk in asserting that this is not the way the Ramones intended it.

In order to both fine tune and expand the approach, a more accurate sentiment analysis could be undertaken paying attention to 5 additional tasks at least:

• in the lyrics, identify the sentiment words preceded or followed by not;
• review and, perhaps, amend the sentiment lexicon(s) to better reflect the punk rock subculture;
• focus on relative more than absolute frequencies of words;
• add terms’ inverse document frequency analysis to measure the impact of the words that are rarely used;
• use ML to spot/predict combinations of n-Grams, topics, writers that would “guarantee” a better ranking in the charts.

The dataset and complete R code of this post can be downloaded from this link.