Searching and Structuring the Twitter Stream for Crisis Response: A Flexible Concept to Support Research and Practice

Kurzfassung

In the context of crisis situations, the high information value of Twitter has already been demonstrated in several studies. However, there are various issues that prevent users from deploying information from social media in practice. This paper outlines the experiences and results of the Data4Human project. In collaboration with the World Food Programme, the German Red Cross and the Humanitarian Open Street Map Team, requirements were defined that have to be fulfilled for a practical use of the data. On this basis, a modular processing system was developed which, in combination with a dashboard, enables the practitioners to evaluate the data and analysis results in a structured and interactive manner. Pre-trained and adjustable machine learning methods combined with search- and unsupervised aggregation capabilities allow to yield different thematic views on the data. By using two use cases, Cyclone Idai and Kenneth in 2019 in Southern Africa and the current war in Ukraine, it becomes apparent, that interactive and customizable methods are of great importance, as well as an effective data representation that enables the identification of relevant content. Based on the results obtained and the user feedback received, future research guidelines are defined in order to close the evident gap between research and practice.