Your Task

Create a swear word cloud which is fit for the Cloud Appreciation Society.

Making it Rain… with Data

We continue to use the profanity based queries we have developed in the following posts (Foulest Mouth of Them All, The Taskmaster’s Potty Mouth).

Some data additional data munging was performed to ensure that all profanities uttered were fully accounted for; namely we have to explode a Python array/list containing profanities through transform and unnest functions.¹

The Overall Swear Forecast

We start with the overall forecast for Series 1 to 16 of UK Taskmaster. As you might assume from the task brief, we will be using the visualisation medium of word clouds to show our appreciation to the swear words uttered.

Word clouds are a simple way to visualise word based datasets. We can capture which words are popular and how frequently they are used (more frequent words are larger in font size), and the variety of words in our dataset.

Figure 1: Overall Taskmaster Swear Cloud

Our findings are in line with external stuies concerning the most common swear words in the UK (Guardian Article):

• Fuck and Shit are the two most commonly used swear words in the UK.
• It is hypothesised that they are frequently used “to emphasis a point in conversation or to build social bonds, rather than with specific intent to cause offence”.
• This supports our Tier4 Swears insights; these swears are used in specific contexts and thus not likely to be used frequently.
  • In my opinion, these swears also have a stronger intention to offend behind them (more than Fuck and Shit at least), and are more directed to an individual or group of people which may also contribute to their low utterance frequency.
  • Despite initial appearances, Taskmaster is less about “offending and insulting people” and more “being entertained and laughing at comedians failures (and successes) in inconsequential tasks”.

The Series Swear Forecast

We have the overall Swear Forecast and insights, but we should also check whether these insights are consistently seen across each series, or if particular series(es) is driving fuck and shit usage.

Figure 2: Taskmaster Swear Cloud by Series

Getting My Head out of the Clouds, and Going to the Bar.

This is my first time generating word clouds and I thought I would share my personal experience with them (so far). I am using the ggwordcloud function from the package of the same name to create these plots. My first attempt at this post also used the wordcloud and wordcloud2 package, before switching to ggwordcloud to stay within the convenient ggplot ecosystem.

I would say that care needs to be taken when creating word clouds. There is a random component when these plots are generated, and certain argument options for ggwordcloud can greatly affect the plots and conclusions we can draw from them, namely the scale argument which represent the (relative) font size of the most and least frequent word.

On many occasions and with certain arguments, words would be excluded from the cloud as they were too big and did not fit on the plotting area. These exclusions would be fine if they were infrequent words, but after some digging around, I later discovered that frequent words (Tier1-2 Swears) were being excluded on some instances. These exclusions are incredibly misleading as to what are the most frequent uttered swear words.

Based on this experience, I would advocate that a barplot² may be the safer option in displaying the same, if not more, information as that on a word cloud.

Figure 3: Barplot of Swear Word Utterances over Taskmaster Series 1 to 16.

Figure 3 is the barplot equivalent of the word cloud presented in Figure 1. Minimal additional arguments were used in generating the barplot compared to the word cloud. By having the utterance frequency explicitly stated on the graph, we can glean that:

• Fuck and Shit are the two most popular swear words (as also concluded from the word cloud), with Fuck being the most favoured profanity.
  • This latter remark was harder to deduce from the word cloud.
• These Tier1 Swears are uttered approximately 6 times more than Tier2 Swears (Hell, Piss, Christ).
  • We would not be able to estimate this magnitude from a word cloud directly.
• We are able to clearly distinguish the leading swear words in the Tiers we have created. For example Hell is the most popular of the Tier2 Swears.
• We are able to reassess and redefine our Tier Swear classification, and which profanities belong to it. For example we can clearly deduce that the following Tiers for the lesser uttered profanities.
  • Tier4 Swears: Cock, Bugger, Bitch, Wank, Tits, Twat.
    • The last three could be their own mid Tier Swears.
  • Tier5 Swears: Pussy, Cunt, Whore, Slut.

But I get it, word clouds are very in vogue and are certainly less intimidating and overwhelming for those who do not enjoy data or numbers. Ultimately, I suspect that word clouds suffer the same fate as pie charts which are notorious for being bad graphs for conveying information (one of many pie chart criticism articles).

But then again, a “Bar Appreciation Society” wasn’t presented as a prize task in the show so it doesn’t fit in with the narrative of this post. I’m sure a Bar (or Pub, Drinking Establishment) Appreciation Society does exist though and may be even more well received by the Taskmaster than the Cloud Appreciation Society.

What Have we Learnt Today?