A machine learning approach to spatial content synthesis from crowd-sourcing platforms (for disaster management)
The recent decade has seen an unprecedented surge in available geo-referenced data. Social media providers, such as twitter (text) or flickr (images), have started to incorporate location into their feeds. This provides an opportunity to harness new sources of spatial information that are cheaper than traditional spatial data sources that typically require specialised hardware and expert users. The biggest challenge for effective exploitation of these new spatial data sources lies in merging spatial information that references the same location but originate from different devices or different communication methods. This project aims to merge all layers of spatially referenced information into one spatial data model. One application of such a system is the area of disaster management, where it would take heterogeneous sources and incorporate these in real time in an app for laypeople. We already see smart phones used in crisis situations, such as the current refugee crisis from Syria. Currently, app users read information manually from apps such as twitter and newspapers, and then manually process these individual pieces of information to make a route based decision. A system as envisioned by this project would immediately provide a content merge of all relevant information in order to assist decision making in critical situations.
Timeframe
Start Date: 1/9/2018
End Date: 31/8/2022
Funding
Fiosraigh New Supervisor scholarship
Lead PI(s)
Dr. Bianca Schoen-Phelan