Researchers mine twitter to map emotions 20140519 p5s7wa.html – Breaking News & Latest Updates 2026
Advertisement
Advertisement
This was published 12 years ago

Researchers mine Twitter to map emotions

AAP

Powered by

Australian researchers will mine millions of tweets to create real-time emotional maps of entire cities and countries.

The CSIRO has unveiled software that can analyse the emotional tone of up to 500 tweets every second - around 27 million per day.

The software looks for any of 600 words connected to specific emotions, from joy and love to surprise, anger, sadness and fear.

Advertisement

It also tracks the time zones connected to tweets to determine their location.

Named "We Feel", the project could provide mental health researchers with mountains of information on the emotional state of entire regions in real time, says Dr Cecile Paris, a CSIRO social media analytics expert.

"It could uncover who is at risk, what our emotions depend on, whether we are different in different regions, how we fluctuate over time."

It could also improve current research methods, which are based on time-intensive surveys with small numbers of people.

Yet the software, created for the Black Dog Institute in tandem with Amazon Web Services, must first be validated.

That means comparing its results with the "Black Dog Index", an ongoing survey of Australian emotional wellbeing.

Dr Paris is confident it will match, buoyed partly by the results of a test run in the aftermath of last week's federal budget: "A spike in anger."

Helen Christensen, executive director of the Black Dog Institute, says the software could provide rapidly the kind of information that would usually take years to collect.

Advertisement

"I don't think anyone yet fully understands how powerful a tool social media can be in the mental health field," she said.

It could reveal "hot spots of despair" and shed light on risk factors contributing to spikes in depression.

On an individual level, it could even lead to automated responses to tweets when language indicates a user might be suicidal.

The software is only compatible with English-language tweets, and can only provide a guess at a user's gender based on the name in their Twitter handle.

Advertisement
Advertisement

That can can be problematic if somebody is named, for instance, Alex or Sam.

Dr Paris says there are plans to finetune the software to provide a better picture of who is doing the tweeting.

* The project can be found at wefeel.csiro.au.

* Readers seeking support and information about suicide prevention can contact Lifeline on 13 11 14 or Suicide Call Back Service 1300 659 467

email icon

Contact us

Share a tip-off, video or photo with us

Most viewed in Australia

More to explore