Community, GHC15

Follow, #hashtag and @mention: Mining Social Media for Disaster Response (Danelle Shah

Number of mobile devices has exceeded human population and this only continues to increase
New apps and platforms
Advances in data science

The challenges faced in using data for disaster recovery
Problems of big data – Volume velocity variety veracity value
Requires processing and sense making of the data
Presence of policies in social networking data

Crowd sourcing has been used very successfully in this field over past few years image video tagging, NLP, .

Much of data collected by noon profit orbs is present in form of PDFs

Geo tagging is gold when trying to map humanitarian crises data. But
Only 1-5% of twitter data is Geo tagged.
A student did Geo estimation of tweet content. User mention Geo location, content based  Geo location

The major part of the talk is about data fusion I.e from different social networks to get better estimations.
They now have a system that identified and trues to contact family of affected people.
There are explicit and implicit connections with social media.
Explicit – names and connections on witte.
When the name doesn’t look like first and last name that’s something machines are not able to figure out full name.
That’s something they are trying to figure out- estimate owner of twitter handle.
Figuring out where a person lives is useful because in case they are asking for hep and not get tagged then this work helps figure out.