Social media activity on flooding (red=high) during a rainstorm in urban environments termed Public Concern Index (PCI). From Wang et al (2020) who proposed these data could be used to inform policies toward flood remediation.
Natural Language Processing (NLP) has been used to map emotions in time (to events) and space (geographical location) for many years.
Bee et al (2018) found it could be useful for early detection of geohazards (floods, earthquakes, volcanic eruptions) to gain real time situational awareness.
With Large Language Models (LLM) all sorts of sentiment can be determined using zero shot classification quite easily. Semi-supervised learning using domain training sets is likely to produce better results however than ‘generically’ trained sentiment classifiers – depending on the use case.
Integrating unstructured text, with temporal and spatial data can lead to new discoveries and areas for further research or action.
References:
Mohanty, S.D. (2021). A multi-modal approach to mining social media data during natural disasters – a case study of Hurricane Irma. International Journal of Disaster Risk Reduction.
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7853661/
Wang, B. et al (2020). Urban resilience from the lens of social media data: Responses to urban flooding in Nanjing, China. Cities 106.
Bee, E.J. et al (2018). GeoSocial: Exploring the usefulness of social media mining in the applied natural geohazard sciences. Janet Watson Meeting 2018.
Early Warning Systems for Natural Hazards
A Geodyssey - Geoscience Text Analytics and Enterprise Search Research 
Leave a comment