EmerGent : Mining social media in large-scale emergencies
One of the biggest sources of Big Data is social media. This represents a continuous feed of unstructured, unvalidated near real-time information about citizens' experience.
When a big city undergoes a crisis, such as a flood, earthquake or riots, masses of valuable information pours into social media sites such as Twitter and Facebook. Currently the emergency services have no established methods of monitoring this data, filtering it and integrating the results into crisis management.
As part of the European EmerGent project, Oxford Computer Consultants are developing cloud-based emergency management services to gather and mine social media for large-scale crises. A service is being developed that can take the stream of social media surrounding a large scale event and transform it into a data source which is relevant, self-consistent and actionable. Social media is already the fourth greatest source of information in a crisis (according to a US Red Cross survey) and is readily available for computing systems to capture and analyse. But how do we quickly process terabytes of data?
How can we parse text with misspellings and in different languages? Can we find false information (as happened in the London riots)? How do we present this so that it is readily understood? How do we maintain citizen confidentiality within privacy legislation?
We are building on our data mining tools and techniques for text-based analysis, and mapping from the REACT project. We are the EmerGent partner responsible for ensuring that all data is used within European Data Protection Laws.
And beyond the project, OCC will be assisting the emergency services in their integration of social media within their incident management and be seeking other opportunities where the technology can benefit Governments and Industry.
The EmerGent project has received funding from the European Union's Seventh Framework Programme for research, technological development and demonstration under grant agreement no 608352.