Geospatial extreme event establishing using social network’s text analytics

Information generated and transmitted over social networks, can help with understanding the development of events that are taking place in reality. With the great increase in usage of smart mobile devices, it is expected to have more data related also to location. Identifying the location of events is of great significance especially in case of disaster or other emergency. Unfortunately, social networks such as ‘Twitter’, ‘Facebook’, and ‘Instagram’, do not always support accessibility to information related to the exact location of neither the user nor the device. For example, in ‘Twitter’, <4 % of the messages carry the precise GPS position. This is due to privacy reasons, device settings and social network regulations. In this study, we use the content of the textual messages to detect the user’s location or, better, the location of the event. Predefined keywords and a set of filters were used to extract and analyze data acquired from the social networks. The analyses of these data focused, in this step, on geographical locations. Findings show that names of places as communicated over the networks corresponded, in many instances, to the actual locations of the events. In the case of major events (in terms of National impact), the volume of messages with location names increase dramatically relative to ‘normal’ periods when there are no ‘extreme’ events. It was also found that 14 % of the messages included the name of the city or even the street name, related to the event.