Social media has changed the way our current society produces, distributes and consumes informational content.
The wide availability of Twitter messages to the general public has cast a favorable light in using this data source for research purpose. In case of natural hazard, tweets can be easily caught to apprehend crisis impacts, people’s feelings or needs. Finding out promptly newsworthy topics can prove by any means useful for crisis management.
A detailed study about automatic content analysis methods did by J. Grimmer, B. Stewart (2013) reviews the state-of-the art in analyzing massive collections of text. Automated content analysis offers a great variety of techniques to measure topics of interest. The paper stresses about the importance of accepting that Quantitative Models of Language should be treated with regards to the specific case of study, as there is no proven universal method that fits all the cases. It also highlights some of the benefits and drawbacks of dictionary methods vs supervised and unsupervised learning that should be taken into account when building a model.
In Snowball, the aim is to build a model to analyze the numerous and spontaneous tweets during a flood, a storm or an earthquake then compute high level indicators. A robust approach to crisis management begins with a strategy prior to the upcoming crisis. It is essential to analyses what are the feelings, needs and more widely topics of interest of the impacted population.
The experimentation of social media integration within traditional crisis management tools should first of all help to define relevant topics and use cases. It should also provide a feedback on techniques, best practices and software architectures for social media analysis.