Analyzing Denzel Curry’s Lyrics Through Text Mining Methods

By Nikhil Sharma and Raghava Govil

UCLA DataRes
May 10 · 10 min read
Rich Fury/Getty Images for Coachella

Note that there is some profanity in this article, included solely for the purpose of lyrical analysis.

Denzel Curry. Denny Cascade. Zeltron 6 Billion. The Black Metal Terrorist.

A man of many names and talents, Denzel Curry is a rapper from Carol City, Florida who has been making some of the most head-banging yet simultaneously nuanced hip-hop out there for a better part of the last decade.

Though most people know him as the guy who rapped fast to kids flipping water bottles, Curry’s albums present an artist with much more musical depth and variation than the memes tied to him would suggest. To bring to light how complete an artist Curry is, we wanted to examine his lyrics with text mining methods to uncover specific lyrical insights.

The data for this project was taken from genius.com. To scrape the lyrics, we used the “genius” package in R, which allows one to get all the lyrics of any song or album for any artist. For our project, we used the lyrics from all of the songs from all five of Curry’s studio releases: Nostalgic 64, 32 Zel/Planet Shrooms, Imperial, 13 and TA13OO.

Word Cloud

A word cloud is a collection of different words used in a document (for us, a song, an album or all his lyrics) with the size of the word corresponding to its frequency: bigger the word, the more frequently it was used. Here is our word cloud for all of Curry’s lyrics from these five projects:

From here, we can see just how profane Curry’s music is, as many of the most frequently used words here are profanities. However, beyond this, many elements of Curry’s personal vernacular pop up throughout his music. For example, there are words such as “ultimate” (which Curry considers himself to be in songs such as “Ultimate” and “ULT”) and “g**k” (not to be confused with the slur, but rather a term used in South Florida that describes a person “not with the in crowd”, as Curry himself describes). Ultimately, this word cloud shows over his five studio releases, Curry stays true to himself by utilizing consistent jargon that makes his songs distinctly his.

Sentiment Analysis

When humans read text, we often try to think of its emotional intent, allowing us to classify the general mood and tone of the writing. Sentiment analysis has the same objective, as it attempts to find the underlying sentiments of a certain chunk of text.

Often, sentiment analysis marks text by using dictionaries. Dictionaries such as AFINN, bing and nrc give either sentiment scores for certain words or categorize words into different sentiments. Using these, we often assign scores or sentiments to each word in a body of text and either take the sum of our sentiments or calculate scores to develop quantitative measures of sentiments. For our project, we used the nrc lexical from the “tidytext” package in R, which is a 2-column dataframe of 6468 unique words and the sentiments that they convey: positive, negative, anger, anticipation, disgust, fear, joy, sadness, surprise, and trust.

To prepare our data for analysis, we implemented a natural language processing concept known as lemmatizing. Lemmatizing involves converting a word to its root word, or ‘lemma’; for example, words such as “playing” and “played” simply become “play”. This makes it easy for our NLP dictionary to limit our analysis to one variation of a given word, since different forms of a word generally carry the same sentiment.

In order to ready our scraped lyrics for sentiment analysis, we parsed the line-by-line lyrics into separate words on each row. Then, we inner-joined our dataframe of lyrics with the nrc lexical. The output was a data frame with each word in the lyrics and its associated sentiment. We condensed the data and added up the counts for sentiments by song, thus giving us sentiment counts for each of the ten sentiments for all songs.

Here are some results for a selected batch of Curry’s songs:

Each number here represents a sentiment count.

You can find a full spreadsheet of sentiment counts for all 55 songs from the five albums here.

The results we have here are interesting, as we can see how some of his most popular songs fare in terms of certain sentiments in the lyrics. However, this table also displays one of the big shortcomings in this method of analysis. According to our analysis, the song “Threatz” ranks as the most angry and one of the most fearful and negative songs in his discography, while also ranking as the most trustworthy and the one of the most positive. This is because in the chorus of the song, Curry and Robb Bank$ repeatedly chant that “[they] don’t take kindly to threats”. While it is clear that the message conveyed here is, well, threatening, our inner-join method cannot recognize that and flags the word “kindly” with the joy, positive and trust emotions from the nrc lexical, since it only tags one word at a time (unigram analysis) rather than considering the context of a word within an entire phrase.

Here are our results over all five of his projects:

Note that 13 has relatively less sentiment counts because the EP is only thirteen minutes long.

Curry’s music generally follows a theme in terms of emotion: lots of angry, fearful, negative music with a few sprinkles of joyous and positive tunes. It is interesting that the emotional distributions are quite similar throughout his albums, considering how different all of his albums are musically. Throughout his career, Curry has gone through phases of experimenting with variants of heavy and psychedelic hip-hop before tying all of these together in his most recent album TA13OO. Still, he has managed to set a specific tone throughout all of his projects that is a signature of his unmistakable style, which is a testament to how gifted a songwriter Curry innately is.


Welcome to the Darker Side of Taboo

TA1300 Album Cover

Last year, Curry released a riveting 13-track personal odyssey in TA13OO. One of the best hip-hop albums of the decade thus far, TA13OO boasts an incredibly wide range of hip-hop sensibilities; from braggadocious, trunk-knocking trap bangers to relatable, introspective ballads, the album features a trove of material that almost any music fan could find enjoyment in.

One interesting aspect that accentuates the variety of material on the album is the concept that Curry focused the album around. He parsed and released the album in three acts: Light, Gray and Dark. The music starts out relatively upbeat and vibrant in the leading Light act. As we progress through the album, the music and lyrics both get increasingly dark and despondent, culminating in a furious storm of an ending. While the album as a whole has a predominantly melancholic tone, it evolves musically and lyrically through the acts to develop its concept.

We thought it would be particularly interesting to analyze this album through the lens of sentiment analysis, since the album is largely centered on a variety of emotions. We wanted to know if analytical methods of text analysis could give us an interesting point of view on the music at hand. Thus, we grouped the songs in TA13OO by act and examined sentiment distributions through the album.

As evidenced by the plot, the Light act has a fairly even spread of emotions throughout. One might be surprised to observe that seemingly negative emotions such as negativity itself, anger and fear match and even outrank seemingly positive emotions such as positivity itself, joy and trust. However, as previously mentioned, the album has undercurrents of gloom throughout. Take the opening track, “Taboo”, for example; though this song opens the Light act, the subject matter is intensely sorrowful, as he calmly sings to a victim of molestation, perhaps in an attempt to cope with some personal demons. Some tracks, such as “Black Balloons”, exemplify the balanced nature of this act of the album. On this song, Curry and GoldLink mull over the depressive thoughts that hang over their heads in the form of black balloons. Though the song deals with many sinister emotions, the ultimate message of the song is one of perseverance:

Soon, black balloons pop / Let it be the day the pain stops.

Following these songs, the Light act features the most bright material on the album. Specifically, the song “Cash Maniac” has the highest sentiment counts on the entire album for anticipation, joy, positivity and trust, which makes sense as this song features Curry frivolously celebrating all of the money he has earned thus far. This continues onto the next track “Sumo”, where Curry transitions to brashly bragging about his wealth while still keeping the mood upbeat on this aggressive hit of a song.

Moving into the Gray side, there is a distinct spike in many negative sentiments. Sentiment counts for disgust, fear and negativity nearly double relative to those in the Light act. In terms of negativity, three of the five most negative songs by count of negative words fall in this act of the album. At the start of the Gray act, we get “Super Saiyan Superman” a distinctively ethereal yet similarly boastful banger to the previous “Sumo”. Then, Curry brings “Switch It Up” and “Mad I Got It”, which both touch on a looming sense of paranoia that pervades through Curry’s entire discography. He fears people whom he perceives as his real friends turning on him on “Switch It Up”, and goes so far as to tell a story of a former friend holding him at gunpoint due to jealousy of his success and riches on “Mad I Got It”. The next two songs paint different shades of hopelessness in the album in the form of social commentary. “Sirens” bluntly addresses the political climate of the country with respect to issues such as police brutality and Donald Trump, while “Clout Cobain” sharply addresses the destructive nature of the clout-chasing circus that many young rappers in the music industry unfortunately find themselves embroiled in. According to sentiment counts, “Sirens” ranks in the top three songs in anger and fear among the whole album, and “Clout Cobain” ranks in the top four songs in negativity.

Finally, we reach the Dark act, which is by far the most vitriolic of the album. Thus far, the album has steadily mounted in dark emotions, which results in the explosion of rage that is this act of the album. Relative to the previous acts, sentiments such as anticipation, joy, positivity and trust are far lower than anger, disgust, fear, negativity, and sadness. The opener to the Dark act “The Blackest Balloon” continues some of the themes of the previous songs with a significantly sharper edge, containing material on Curry’s late brother haunting him and the untimely death of Lil Peep. “Percs” builds on these previous tracks from a unique angle, as Curry furiously differentiates himself from his peers, specifically in his opposition to recreational use of pills such as percocets and to unintelligible, mumbled flows. After these tracks, we reach the nastiest song on the whole album: “Vengeance”. By sentiment counts, this song leads the album in anger, disgust, fear, negativity and sadness; quite the violent song, as the title suggests. This song definitely fits the bill that our sentiment analysis tags it with, as Curry, JPEGMAFIA and ZillaKami all deliver murderous verses with junkyard flows that culminate into a beautifully vicious package, matched with an equally enjoyably reprehensible video. Finally, the album ends on “Black Metal Terrorist”, an absolute barrage of aggressive rhymes revolving around Curry trailblazing his path of greatness through other rappers.

Each number here represents a sentiment count.

Overall, we feel that our sentiment analysis confirmed and enhanced our understanding of the concept of Curry’s album. As the acts of the albums got darker, so did the tone of the music, as we can confirm both by listening closely and our sentiment analysis. Unfortunately, there were some difficulties that we did our best to mitigate but were still prevalent. As mentioned before, unigram analysis often does not account for the context of a single word within the frame of a sentence. In the nrc lexical, the word “cash” is tagged as anger, anticipation, fear, joy, positivity and trust. The problem here, in the framework of our project, was that it initially classified “Cash Maniac”, unequivocally the most jubilant song on the album, as an angry and fearful one because of the sheer number of times the word “cash” is used in the song. Though we were able to spot this and alter the sentiment counts for that specific song, similar issues might have occurred in our analysis without our knowledge.

Thus, a next step might be to consider analyzing the lyrics using n-grams to gauge our sentiments more accurately. Still, we have something quite fascinating here, as even a relatively rudimentary analysis of the data fully supports our ear-test as music fans, showing us the power of sentiment analysis.

One of the punchier lines on TA1300 comes from “Switch It Up”, where Curry says:

They only know Denzel Curry / But they really don’t know Denzel!

Hopefully, we now have a better understanding of who he is, and how he works with and incorporates certain emotions in his music.


If you would like to see our code, you can check it out here.

We would like to acknowledge two sources which helped us immensely:

UCLA DataRes

Written by

UCLA’s first Data Science Club. datares.github.io/, facebook.com/ucladatares, linkedin.com/company/ucla-datares

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade