Creating Spatiotemporal Semantic Maps from Web Text Documents - Space-Time Integration in Geography and GIScience - page 162

Fig. 10.4 Locations extracted from news reports about tornadoes in April 2012. Temporal
information refines the spatial pattern of locations affected by the tornadoes associating events
with more days as compared to the map in Fig. 10.3

mentioned in the documents (e.g., Kansas, western Kansas, northern Oklahoma,
Nebraska, Michigan, Wisconsin, Minnesota, Iowa and Louisiana) are represented
on the map with a unique symbol to highlight the uncertainty of the locations of

these events.
When temporal information about the tornado events is extracted and combined
with spatial information, some of the locations identified in Fig. 10.3 where only the
document date is used are actually associated with dates different to the document
date and a greater range of dates is represented (Fig. 10.4 ). For example, Norman,
Oklahoma, extracted from the news report on April 14, is associated with events on
April 13th. Some state names, such as Kansas, Iowa, Nebraska, Oklahoma, reported
on April 15th, are revealed to be actually associated with events on April 14th.
It's also possible to detect some locations that are associated with future temporal
information. After extracting and processing temporal references extracted from
each document and using these in conjunction with spatial references, the locations
displayed on the map represent a temporal sequence of events that is more realistic
and meaningful. In this way, it is possible to distinguish information about current
or past movements vs. future movements.