General Conclusion - Geographical Information Retrieval in Textual Corpora - page 103

4.2.3. Expansion of the vocabulary for a qualitative representation of the
geographic dimensions
We are inspired by the Geooreka! system described in [BUS 09b]. This system
is devoted to the spatial and thematic IR on the Web. Buscaldi and Rosso [BUS 09b]
proposed representing a spatial search criteria by a set of toponymic names, which
they combine with the other terms of the search and submit to the yahoo! engine.
Following the same principle, we hypothesize that this approach could be
extended to the spatial, temporal and thematic criteria. Figure 4.10 describes the
main steps relative to the process of user need definition. The spatial criteria are
described graphically on a map. The denoted geographic zone serves as a parameter
in the retrieval of corresponding toponyms. One or more gazetteers are queried in
order to extract the toponyms relative to the zoom level and the geographic zone
defined in the previous step. For a geometry covering the commune of Laruns, for
example, a knowledge base such as a gazetteer would give us all the oronyms,
hydronyms as well as points of interest such as huts, mills and fortifications. A new
filtering of the obtained toponyms as a function of the user's profile or other
arguments of the query can be considered.
The temporal criteria are also described graphically on a time line. The period
framed in such a way serves as a parameter in the retrieval of dates, hours,
praxonyms (historical events, diseases, cultural events) or phenonyms (hurricanes,
high/low pressure zones, stars, comets), for example. One or more gazetteers are
queried in order to extract the terms relative to the zoom level and the period defined
in the previous step. For a period covering the year 1804, for example, a gazetteer
would give us the year 1804 but also the Coronation of Napoleon, the Empire, the
Civil Code, etc. A new filtering of the obtained terms depending on the user's profile
or other arguments of the query can be considered.
This approach, which is very pragmatic, can be especially interesting in such
contexts of use for which space, time and theme are well bounded and described in
gazetteers that may be used in the phase of definition of the user need. For a sports
reporter specialized in rugby, for example, the spatial gazetteer could be used for the
description of the different playing areas of rugby terrain, a temporal gazetteer would
describe the periods of game relative to a match and finally, a thematic gazetteer
would be devoted to the rugby-specific vocabulary. Likewise, in another framework
of use, the spatial gazetteer could describe the rugby world (hemispheres, countries,
provinces, clubs, etc.) and the temporal gazetteer could describe the dates and events
that made rugby history, for example.
We can thus imagine ad hoc gazetteers describing a building, a set of buildings,
opening times, manufacturing times, sale periods and specific domain vocabularies.
The qualitative representation of the initial version of the query is then submitted to a
classic search engine.