Theresa, Trump and textual analysis
Perfect to tweet: 83% of Trump’s inauguration speech sentences were 140 characters or fewer
Geoff Copps, head of research at Mediabrands Marketing Sciences, explains how fascinating consumer insight can be gleaned from using different types of textual analysis
As Prime Minister Theresa May prepares for her first encounter with President Donald Trump, she must be wondering what to expect. Her aides will be poring over Mr Trump’s speeches and online comments in an effort to glean fresh insight into who he is, what he is thinking, and how he will react to upcoming events.
These three questions also preoccupy marketers as they strive to better understand their target audiences.
Over the past decade, online communications have provided us with plenty of material to use in this context. By applying various types of textual analysis to ‘texts’ such as tweets and blog posts, we are able to enhance our understanding of consumers and other influential voices in the marketplace.
This can happen in a variety of ways. We can use it to understand the power of brands in online consumers’ lives, and to gauge the impact of advertising campaigns. Or we can use it to inform influencer strategies by identifying the most suitable online voices through which to amplify a clients’ brand message.
In a marketing context, the type of textual analysis that gets most attention is semantic – i.e. that branch of analysis concerning the interpretation of specific vocabulary used by consumers. Viewed in isolation, semantic analysis can have limitations. Vocabulary is difficult to interpret reliably.
Even the most advanced natural language processing and machine-learning techniques may struggle with some specialist language usage. Irony and sarcasm are cases in point; they happen also to be especially common online. The meaning of a word will also vary by demographic group (consider how ‘wicked’, ‘cool’ and ‘sick’ are used differently across the generations).
But by combining semantic techniques with a more quantitative approach focusing on syntax and other parts of speech that tend otherwise to get overlooked, we can start to get somewhere. In its emphasis on numbers, this complementary approach can seem ugly and inelegant at first; but it can be invaluable for capturing the wider context and providing a ‘macro-view’.
Let’s return to where we started by considering 2017’s most high-profile piece of comms: Mr Trump’s inauguration speech.
In his speech Trump was considerably less likely to use the indefinite article (‘a’, ‘an’) than Obama in 2009 – implying that there are more certainties in Trump’s worldview.”
Setting aside straight political commentary, much has been written on the semantic front – probing phrases such as ‘American carnage’, and seeing in the usage of ‘America’ and ‘American’ evidence of a pivot in policy-making.
So what can a more syntactically focused, quantitative approach tell us?
Consider the most obvious statistic, word count. Trump’s speech, clocking in at 1,433 words, was relatively short compared with previous inauguration speeches.
Things get more interesting at the sentence level. Trump’s speech comprised 87 sentences, with the average sentence length being 94 characters – making the majority of what he said tweetable. 83% of his sentences were 140 characters or fewer.
Compare this statistic with former President Barack Obama’s 2009 inaugural address, in which the equivalent figure was below 70% (excluding conventional periods such as ‘Thank you’).
All of this seems to confirm the breeziness of manner, sloganeering and – to misquote a creative agency – brutal simplicity of thought that were Trump’s ace cards during his campaign. He looks set to continue in this vein.
Revealingly, Trump also has an above-average fondness for the connector word ‘and’. He often uses it at the beginning of a sentence, and he is not afraid of repeating the word. This creates a cumulative effect, a kind of heaping up of phrases which makes his supporters feel well catered for (“We will build new roads, and highways, and bridges, and airports, and tunnels, and railways all across our wonderful nation” being a good example).
By contrast, Obama was more likely than Trump to use other connector words such as ‘or’ and ‘but’, reflecting his more sophisticated and deliberative – critics might say too deliberative – style.
In his speech Trump was also considerably less likely to use the indefinite article (‘a’, ‘an’) than Obama in 2009 – implying perhaps that there are more certainties in Trump’s worldview.
However, the biggest difference between the two men’s first inaugural addresses was in the use of future-tense verb forms in the first person, ‘shall’ and ‘will’. Trump used ‘will’ 40 times, most notably in a pattern of repetition toward the end of his speech. In 2009, Obama used ‘will’ and ‘shall’ a total of 20 times – half as many in a speech that was almost twice as long.
Textual analysis of different types, then, can reveal useful insight about a speaker or writer. Extra-semantic factors such as sentence length and repetition can contribute to our understanding of how an individual or group of individuals is thinking and feeling.
To gain the fullest insight into a consumer or influencer, you need to look not only at what they say but also at the way they say it. This isn’t a new idea, of course. The medium is the message – and few are better placed to take that point than media agencies.