14 January 2019

Language patterns in politics in 2018

Jump to infographic ↓

A lot of what drives Lexico is the view that computers should be able to understand us when we talk to them (if only at a basic level).

That's a pretty understandable objective, at least in the context of your day-to-day work.

But language as a concept occupies a much more important position in broader society.

Without it, could we have forged strong democratic institutions? Collaborated across continents to drive the sciences forward? Built global giants like Google?

We can all admit that we wouldn't have, given how intrinsic language is to human civilization.

The problem we're facing now is that there's an absolute glut of it. It's increasingly difficult to separate fact from fiction, truth from rhetoric.

There's no doubt that language has always been a tool for manipulation - but what Cambridge Analytica exposed was the potential for targeted manipulation on a mass scale.

Making sense of the torrent of data thrown up on a daily basis can be difficult - and even more so when it comes to politics.

That isn't necessarily new - which is why we historically relied on the media to do it for us. But what do we do if we feel the media is pushing its own unreliable narrative?

Question time at the Australian Parliament is a perfect example. Every year, thousands of pages of transcripts are produced, evidencing attempts by government and opposition alike to out-do one another.

It's too much for the average individual to realistically wade through. Staying reliably informed means we need to be able to analyse patterns from large, unstructured data sets.

That's the great advantage that computers have over humans - the size of the input data doesn't really matter. For us, figuring out 1000 sentences needs 1000 times as long as figuring out a single sentence. For a computer, the additional overhead is (usually) trivial.

If we wanted to paint a picture of trends or slogans during 2018, we wouldn't glean it from individual sessions. We'd need to look at the whole year, and we're going to need some kind of program to help us do it. Using some basic word counting techniques gives us some pretty good insight into what went on in 2018.

The source code is here for anyone wanting to peruse themselves - but in basic terms, we extracted all two- and three-word phrases, then sorted by frequency. The idea was to identify a common set of each party's rehearsed catchphrases/slogans, but this actually ended up giving a pretty good overview of the political topics of 2018.

At a high-level, we can see that the biggest topics of contention in 2018 were Taxation, Economy, Budget, Energy and National Security (as officially classified by Hansard, the parliamentary transcription service).

Top 10 Question Time Topics in 2018

Topic Number of Q&A % of total Q&A
Energy 274 12.02
Taxation 179 7.85
Economy 138 6.05
Budget 129 5.66
Workplace Relations 92 4.04
National Security 86 3.77
Schools 66 2.89
Deputy Prime Minister 50 2.19
Infrastructure 44 1.93
Income Tax 37 1.62

Taxation, Budget and Economy

Unsurprisingly, ‘tax cut’ headed up the list of most common phrases during questions for this topic, uttered more than 500 times during the year. Other repeated phrases included ‘family business’ (59 appearances), ‘more jobs’ (57) and ‘tax relief’ (50).

Resounding catchcries like ‘strong economy’, ‘more jobs’, ‘economic growth’, ‘Australian family’ and ‘jobs and growth’ all featured heavily. Both parties sought to appeal to the economic needs of ‘small business’, which was the most-used phrase across all topics (573 appearances total).


Energy was a big topic during 2018, at around 12% of the total questions and answers for the year.

Surprisingly, the discussion wasn't dominated by technical issues or Elon Musk (though ‘South Australia’ featured regularly).

Rather, both political parties hammered the phrases ‘electricity price’ (250 mentions), ‘power price’ (162), along with typical slogans like ‘affordable and reliable’ (104), ‘Australian family’ (83) and ‘blue collar worker’ (64). These show a pretty clear attempt to identify with working-class Australians’ concerns about rising electricity prices.

National Security

If you've lived in Australia for more than a month or two, you'll know the current government prides itself on its tough stance on border controls.

This was reflected during parliamentary question time, with repetition of ‘border protection’ (137 mentions) and ‘operation sovereign border’ (22) emphasising the need to ‘keep Australians safe’ (17).

The government went to considerable effort to paint its policy as humanitarian - regularly linking border policy with the phrases ‘drown at sea’ (48 mentions) and ‘death at sea’ (18).

Wrap up

This is all fairly rudimentary analysis - much simpler than the language models used inside Lexico to understand what you say and bridge it with your application APIs.

But with language becoming increasingly rehearsed, targeted and weaponised, it's important to keep a critical eye on what's being said to us.