Local Knowledge and Social Sensors: Integrated Models of Text Analysis for Disaster Response

The present doctoral research investigates the role of local knowledge in supporting disaster response, by applying cognitive, predictive and ontological models to the study of text message exchange in relevant Social Networks. The lack of studies on local knowledge in risk domains, along with the growing attention of scholars and decision makers to open governance processes, underpin the three main research questions that have been addressed. As for the relevance of local knowledge to disaster response, human communication – in real or simulated emergency situations – seems to be imbued with culturally mediated understandings of spatiality, relationality and actions. Innovative spatial data science tools are therefore needed for tacit and vernacular knowledge to be adequately modelled and operationalized. With respect to the use of text messages in disaster response, it appears that distributed systems (which combine crowdsourcing methods with collaborative hypertext editors) may effectively complement volunteered or public participation GIS – to harness the potential of all-purpose social networks to reach out to the wider internet community. Finally, to foster the interpretation of text messages, three separate taxonomies (regarding Spatial Location, Needs and Actors) – each being linked to a terminal entity in DOLCE foundational ontology –, helped develop a shared conceptualization of risk. Future developments of the present work could concern the further integration between machine learning and ontological models, and advancing text classification methods.