Multiple studies have proven that cell phone data can be very useful in determining the distribution of resources for disaster relief although many privacy issues arise with the use of location data. To collect location data without infringing privacy our model utilizes the information shared on social media to deduce the location data needed in a disaster scenario. It accomplishes this by utilizing the machine learning capabilities of IBM Watson.
Obstacles: Problem: Tweet locations aren’t always real, ex: “Uncanny Valley” “Milky Way” “Hell” Solution: Compare locations to real Google Maps locations Problem: Comparison to Google Maps locations is high latency Result: Improved latency marginally, future iterations would be optimized in order to implement streaming, also attempted quicker but less accurate method involving string filtering
This model can be expanded for use with other various social media platforms and specialized for specific disasters. Latency could be improved to enable streaming.