Designing a tool that maps global and local updates on crisis situations — The case of Lombok
The effects of the earthquakes that struck Lombok at the end of July and beginning of August are devastating. An estimated 157,000 people have been affected. As is often the case in such critical situations, the exact death toll remains uncertain and is expected to increase. It is currently estimated to lie between 131 (according to national institutions) and 347 (according to local institutions). Tourists are leaving the island as soon as they get the chance.
(Inter)national media update people with Internet access worldwide about the general situation in Lombok. As a technical preparation of a future building block of the gaia platform, the gaia team has developed a tool based on A.I. techniques that maps messages (tweets on Twitter, such as the example shown in Image 1) and online articles, and the underlying sentiment of these. The goal of the tool is multifold. First of all, it allows the team to analyze which techniques contribute most to optimized bundling of information, in order to potentially respond better to local needs and demand. Secondly, the cluster algorithms of the tool can be used to cluster any type of information, and could prove useful in later stages of the development of the platform’s technical building blocks. For example, projects launched on the platform could be clustered and ordered by topic. Lastly, a more elaborated version of the tool could be used to analyze the extent to which news articles or messages from official institutions match the real-time information provided by local people who are affected by the disaster.
The current tool scans Twitter and online news sites according to specified search terms, which include “Lombok”, “2018”, “earthquake”, “victims”, “casualties”, “donations”, “aftermath”, “humanitarian aid”, “aid organizations”, “destruction”. A cluster algorithm subsequently clusters tweets and articles, which results in a set of topics. The tool is designed to collect tweets with and without geo location. Image 2 shows a map of the tweets and articles with geo location. The tool allows its user to hover over the map and shows related tweets in the process. In addition, the sentiment (positive, neutral or negative) of these tweets and articles are plotted (see Image 3). ‘Positive’ sentiments are linked to constructive messages (e.g. “Let’s help Lombok by donating”), whereas ‘negative’ sentiments concern messages such as “Tourists ‘forced to pay to board rescue ships’ amid deadly Indonesia earthquake”, as shown in Image 3.
