By Associate Professor Rodney Clarke

One of the problems of using social media during emergencies is being able to assess the various kinds of information that can be in them.

Of the 50,000 tweets harvested during the Queensland Floods of 2011, about half of those that used the official #qldfloods hash tag were retweets, and so provided no new semantic information. Almost none of the contributors had the geolocation feature of their mobile phones enabled. This limits the utility of the message because the tweet cannot be accurately associated with a geographical position. At best this corpus consists almost exclusively of ‘unconfirmed tweets’. A surprisingly large proportion of the remaining tweets had little to do with the particulars of the flood. Many Twitter users were simply in the habit of using currently trending hash tags to broadcast as widely as possible their opinions concerning the government, celebrities and the kind of random thoughts and vents that are often posted on social media sites.

The penetration of mobile devices and social media is increasing at an extraordinary rate. PetaJakarta.org has collected enormous numbers of flood related tweets (those including #banjir or ‘flood’ in Bahasa Indonesia) during the 2015 January monsoon season in Jakarta alone. But if we want to separate critical flood-related tweets from those that are not, then techniques need to be applied or developed to distinguish between them in the deluge of potentially relevant social media messages.

These techniques will be discussed in my next blog post.

Associate Professor Rodney Clarke is Manager of the Collaboration Laboratory (Co-Lab) at SMART Infrastructure Facility.