Major points to hit on

Twitter Habits Change During Emergency

How does this affect the data? Locations may be off. A before location may be determined as place like a park where the Twitterer tweets every morning with their dog on a walk. If this park is in an evacuation zone, then this user comes up as high risk.

During the storm, however, they do tweet from their home, because they are stuck there. So the only tweets they ever send from home are during the storm. Classifying this type of behavior is troublesome because it looks different than it actually is.

Frankenstorm Apocalypse

A few social geo check-in sites, mainly Swarm and Foursquare had specific events for Hurricane Sandy in which many users checked in. When shared on Twitter, however, there is little we can filter out as the URL and text is always changing. Filtering out all Foursquare tweets would be foolish because much of the rich geo data is specifically from Foursquare.

Completeness of the Data

  1. Need to ensure that we have the full contextual streams, not just limited to the