Social networks have been used to overcome the problem of incomplete official data, and provide a more detailed description of a disaster. However, the filtering of relevant messages on-the-fly remains
challenging due to the large amount of misleading, outdated or inaccurate information. Extending earlier work a new paper presented this week at Geoinfo 2015 presents an approach for the automated geographic prioritization of social networks messages for flood risk management based on sensor data streams.
It was evaluated using data from Twitter and monitoring agencies of different countries.
The results revealed that the proposed approach has a potential to identify valuable flood-related messages in near real-time.
Future work lines should take account of using the prioritization of social network messages as one step to further filtering and classifying the quality of crowdsourcing. Besides that, it can serve as basis to improve machine learning models that consider geographical links.
ASSIS, L. F. F. G., HERFORT, B., STEIGER, E., HORITA, F. E. A., ALBUQUERQUE, J. P. (2015). Geographical prioritization of social network messages in near real-time using sensor data streams: an application to floods. XVI Brazilian Symposium on Geoinformatics (GEOINFO). Campos do Jordão, SP, Brazil.
http://www.geog.uni-heidelberg.de/gis/publikationen_conference_en.html