Black Dog tracks mental health through the emohose
Researchers from the Black Dog Institute and the CSIRO have developed a tool for measuring and monitoring emotional states using real-time data from Twitter, which they hope to use to better understand regional fluctuations in mental health.
The We Feel tool allows researchers to create a visual representation of emotional content and compare it across space and time. The researchers say that the power of Twitter is its immediacy, allowing them to track changes in emotional states in real-time rather than months or years after the event.
They have also released a REST-ful API to allow other researchers and developers to access the raw data behind We Feel. The data is shared under a CSIRO data license so developers can use it royalty-free for non-commercial purposes.
Black Dog Institute director Helen Christensen said that being able to identify and analyse this kind of data in real time was of enormous benefit to both research and public health.
"Currently, mental health researchers and associated public health programs use population data that can be over five years old,” Professor Christensen said.
“Should the real-time data gained using this incredible tool prove accurate, we will have the unique opportunity to monitor the emotional state of people across different geographical areas and ultimately predict when and where potentially life-saving services are required."
The project is being supported by Amazon Web Services' Kinesis platform to deal with the huge data volumes of Twitter, and by Gnip, a commercial reseller of Twitter data that can provide access to every publicly available tweet dating back to Twitter's launch on March 21, 2006.
The researchers are using Twitter data rather than other social media platforms such as Facebook as Twitter is a public platform and Facebook users commonly apply privacy settings. The developers says that the tool has been specifically designed to only look at the big picture and not to identify individuals.
The researchers say that of the hundreds of millions of tweets posted each day, a large part concern their emotional state.
“We Feel is about tapping that signal to better understand the prevalence and drivers of emotions,” they say. “We hope it can uncover, for example, where people are most at risk of depression and how the mood and emotions of an area/region fluctuate over time.
“It could also help understand questions such as how strongly our emotions depend on social, economic and environmental factors such as the weather, time of day, day of the week, news of a major disaster or a downturn in the economy.”
Tracking down exactly where these tweets are emanating from is tricky as many people don't specify where they are tweeting from or use mock locations such as the moon, so the researchers are tracking the tweets by time zone. This works well for Australia, which is the focus of their research.
Professor Christensen said that as an example, her team analysed the emotions of Australians before and after the budget announcement last Tuesday.
“As Minister Hockey rose to speak, there was a sharp spike in fearful tweets,” she said. “As he finished, there was a surge of angry tweets, significantly larger than seen the night before. We also picked up increases in sadness and a distinct reduction in joy."
The tool gathers its tweets from three “hoses” or sources. Twitter's public API is known as the gardenhose, and Gnip calls its Twitter stream capability the “firehose”, so the researchers have taken up this terminology to develop an “emohose”.
A random sample of one per cent of the tweets is taken from the gardenhose, 10 per cent is from Gnip's firehose, renamed the “decahose”, with the main source the “emohose”, which will monitor a large sample of emotional vocabulary derived from Parrott's tree-structured list of emotions.
The researchers say gardenhose delivers about 900,000 English tweets a day, of which about 250,000 contain emotion terms, and the decahose delivers 10 times that amount. “The emohose delivers about 27 million English tweets a day, all of which contain emotion terms,” they say. That averages out to 19,000 tweets per minute.
Along with the visual explorer, We Feel provides a table builder and a REST API that researchers and developers can access.
Posted in Australian eHealth