Creating Spatiotemporal Semantic Maps from Web Text Documents - Space-Time Integration in Geography and GIScience - page 154

on their interests, the possibility of automatically tracking spatiotemporal and
semantic change relating to events is a more recent challenge for IR, one that will
improve understanding of event-based information (e.g., hazard events) described
in collections of unstructured web documents from spatiotemporal and semantic
perspectives. As the techniques of IR have improved over the past decade, IR has
begun to accommodate and exploit a broader range of information, for example,
geographic information. Geographic information retrieval (GIR) stems from the
discipline of IR, and involves not only many IR methods, such as indexing,
searching, browsing and querying documents, but also considers the geographic
scope of documents (Kemp et al. 2007 ; Jones and Purves 2008 ; Keller et al. 2008 ;
Teitler et al. 2008 ). With GIR, natural language processing (NLP) techniques are
used to extract geo-references such as zipcodes, addresses, well-known landmarks,
toponyms, telephone and area codes from web pages (Purves and Jones 2011 ).
In geographic information science (GIScience), modeling geographic dynam-
ics based on spatiotemporal information extracted from the web is a growing
field of research. Geographic dynamics refers to the change or movement of an
event with spatial and temporal characteristics, and involves understanding “the
fundamental functions of relevant forces and their interrelationships in space and
time” (Yuan and Stewart Hornsby 2008 , p. 7). The topic of modeling dynamics of
geographic domains includes: calculating the characteristic patterns of movement
of individuals or groups (Laube et al. 2007 ; Dodge et al. 2008 ; Yuan and Stewart
Hornsby 2008 ; Wood and Galton 2010 ; Dodge et al. 2012 ); time geography analyses
that investigate patterns of people's activities from spatial and temporal perspectives
(Miller 1991 ;Kwan 2000 ; Raubal et al. 2004 ;Yu 2006 ;Shawetal. 2008 ;Chen
and Kwan 2012 ); and modeling the trajectories of moving objects such as people,
vehicles or natural phenomena (Dodge et al. 2008 ; Stewart Hornsby and Li 2009 ;
Demsar and Virrantaus 2010 ). Automatically capturing spatiotemporal and semantic
information from the web and representing the underlying relationships through
geographic information systems (GIS) not only transfers text content to a visual
representation that preserves informational characteristics from the documents, but
also offers people a better understanding of spatial and temporal characteristics that
are described in the immense resource of web documents and provides a clearer
overview of dynamic change patterns and trends over space-time.
In this work we discuss a hybrid method of combining the result of natu-
ral language processing techniques and ontologies to improve the extraction of
semantics of hazard-related events sourced from web texts. This work combines
principles from the field of NLP and GIR with ongoing work in the field of
temporal GIS where there is an interest in modeling geographic dynamics. For this
research, an open source software tool, General Architecture for Text Engineering
(GATE) ( http://gate.ac.uk/ ) , is used as the primary tool for spatiotemporal and
semantic information extraction. The motivation for this research is to automatically
represent the spatiotemporal information and track moving entities over space-time
to show the spatial extent and trends of the dynamics of hazard-related events.
Also, semantics are automatically processed to make the extracted spatiotemporal