Show transcript

[Text appears on screen: Can an algorithm sense our emotions?]

[Music plays and image changes to show Stephen]

Stephen: My name is Stephen and I’m a research scientist. My area of science is called computational linguistics and in computational linguistics what we like to do is we try to put together technology to analyse language that we might hear, but the kind of language that I look at is written language, particularly I look at analysing social media text, so the kind of language you see on Twitter and blogs.

[Image changes to show Stephen with colleagues and then changes to show Stephen seated at a desk and working on a computer]

It’s not just about the minutiae of people talking about what they had for lunch; people are sharing really interesting things.

[Image changes to show Stephen using his mobile phone and then changes to show Stephen standing and watching different algorithms pass by on a screen]

And if you take all that information and you pull it all together, that public data, then what seems to bubble up is the way the community as a whole thinks about a particular topic.

[Image changes to show Stephen selecting a book from a bookcase and then changes back to show Stephen seated and working at a computer]

For example, we worked with the State Library of New South Wales, and in fact this was a really interesting application, because our tools, we had designed them to work for media and communication staff, but the State Library realised that we could use these tools to collect ephemeral content about events like the recent New South Wales State Election.

[Image changes to show Stephen selecting a book from a bookcase and then shows him flicking through the book]

For example, we can see on Twitter what election issues really resonated with that community and that’s a really interesting piece of information for the historians in the future.

[Image changes to show Stephen participating in a game of capoeira]

One of the things I like to do outside of my research life is to play capoeira. Capoeira is a Brazilian martial arts that involves, not just, the physical element but also music and community.

[Image changes to show Stephen walking and talking with another male and then moves back to Stephen practicing capoeira]

But what I find similar between my research, music and capoeira is that there is an element of creating and saying something, so it’s playing music, or creating a sequence of moves that is related to what other people are doing, and it’s the sense of improvisation and communication that I think is an underlying thread in these three areas.

[Image changes to show a head shot of Stephen]

[CSIRO logo appears with text: Find out more]

Show transcript

Meet Stephen

Can an algorithm predict what issues will be important in an upcoming election? Can analyses of social media help us understand how to provide better mental health services for the community?

As we start to wonder what helpful insights might be gleaned from the millions of public opinions, likes and re-tweets shared online each minute, computational linguists like Stephen Wan are already two steps ahead.

How does your work as a computational linguist studying what people say on social media contribute to our understanding of society?

Every minute millions of people share content on their social media – articles, videos, photos, personal opinions and more.

It is a vast sea of data that grows at an overwhelming rate, so much so that it is impossible to read and categorise by hand, yet it may have the potential to provide game-changing insights. These insights could help us understand, for example, social factors that contribute to health issues like depression and obesity, which are often discussed online, to understand how we can improve our health services.

My job is to create algorithms that can turn the bottomless sea of social media data into information that is useful. I develop programs that are sensitive to the subtle nuances of language to help us to analyse billions of tweets; comprehensively sifting through data in a fraction of the time that it would take a person or a team of people!

How can you use computational linguistics in a way that can affect a large group of people, for example an entire city or state?

We recently helped the State Library of New South Wales to collect public discussions on the NSW state election so we could understand why people would vote the way that they did.

Where usually we would have to wait for someone to write articles discussing the public opinion on election matters, our program was able to capture this collective voice in the discussion in real time.

In the context of the NSW elections, what difference was computational linguistics able to make?

Our program could show us the information (news links) that were being shared and with what frequency, highlighting the different topics and their importance to us as a society.

Most importantly we could track this information over time, watching how opinions and media interacted and affected each other, ultimately deepening our understanding of what has an effect on voting behaviour.

What are the unique challenges facing computational linguistics that you’re excited about overcoming?

One of the things that makes language so interesting is the way in which we are creative with it – a word can have a number of different meanings, and this may depend on factors such as the age, gender, location and cultural context of the person using it.

This presents us with a unique challenge: to design automated programs that grasp those subtleties. For example, how do you write a program to detect something as complex and evolving as human sentiment? These are the kinds of questions that we work towards answering each and every day.

Facts & figures

Our digital world

  • Over 80 per cent of all Australian households have internet access
  • 11 million Australians made an e-commerce transaction in 2014
  • Forbes Magazine labelled data scientists the ‘the sexiest job of the 21st Century’
  • 13 million Australians spend over 18 hours a day online.

Social media in Australia

  • Our 14 million Facebook users visit the site on average 16 times a week
  • One minute out of every five (3.6 hours) is spent on social media
  • 20 per cent of our 1.1 million regular twitter users engage multiple times a day.

Find out more

Ask Stephen a question or get the very latest news, stories and breakthroughs that will help you, your family and Australia.


Meet the seven


It takes creativity to ask the interesting questions and it takes innovation, science and technology to answer them.

CSIRO is where creativity meets innovation, developing solutions to make a difference to you, our economy and the planet.

We imagine. We collaborate. We innovate.

More about us