Geography Reference
In-Depth Information
are published. In this way, document date can be used to give a basic temporal
order to the extracted spatial locations. However, as presented in the following
discussions, this information can be improved further when additional processing
is applied to text information.
For this study, we are particularly interested in how spatial and temporal
references available in text work together to inform us about the dynamics of events.
The extraction of spatiotemporal information is accomplished by creating spatial
and temporal gazetteers for text matching, text processing, and geovisualization.
The default spatial gazetteer for GATE contains general world references plus
mostly UK-based places, and so an additional gazetteer using data from StreetMap
Premium for ArcGIS and the US Gazetteer Files for the 2010 Census ( https://www.
census.gov/geo/maps-data/data/gazetteer2010.html ) has been created to support the
extraction of places for articles describing events in the United States. GATE
also provides basic annotation support for temporal references including common
temporal references, such as day, week, month, and year. Documents also often
include additional temporal detail that GATE's default gazetteer alone does not
match, and that enriches our understanding of the event's dynamics. Therefore, we
have extended the default temporal gazetteer by adding a new temporal gazetteer
with additional temporal references, such as early Sunday morning or late night
to provide additional temporal annotation capabilities. The second step of the
extraction process involves the parsing and annotation of spatial and temporal terms,
combining them in a meaningful way. The final step is to export the sorted results.
In this way, the data are ready for geocoding, and mapping and visualization of the
extracted spatiotemporal information in ArcGIS.
Combining spatiotemporal information in a meaningful way is an important
objective for this research. Spatiotemporal data are extracted to capture the salient
details of spatiotemporal dynamics in articles. For many documents, geographical
references are associated with temporal expressions. For example, More than 1,000
volunteers rushed to fill sandbags Wednesday as many in Fargo tried to protect
themselves from a historic flood that is expected to swamp the area . In this example,
Fargo is associated with the temporal expression Wednesday, affording both spatial
and temporal details about an event of interest. However, not every sentence will
have both spatial and temporal information. Five possible cases can arise with
respect to spatiotemporal information in a sentence (Stewart Hornsby and Wang
2010 )(Fig. 10.2 ):
￿
When only spatial information is present in a sentence
￿
Sentences with one spatial term and one temporal reference
￿
Sentences with one spatial term and multiple temporal references
￿
Sentences with multiple spatial terms and a single temporal reference
￿
Sentences with multiple spatial and multiple temporal references
These cases provide a basis for automatically combining spatiotemporal informa-
tion retrieved from text documents in order to use that information to model dynamic
entities in a GIS. It is also possible for text documents to present only spatial
information and not contain any temporal information. In these cases, the document
Search WWH ::




Custom Search