Access by Geographic Content to Textual Corpora: What Orientations? - Geographical Information Retrieval in Textual Corpora - page 8

Figure 1.3. Our GIR study and the main research fields of interest involved -
extended extract of Julien Lesbegueries [LES 07]
In Figure 1.3, we also show the necessity of having spatial and temporal models
for the description of information extracted from texts. A number of processes,
supported by geographic information systems (GIS), for example, dedicated to
spatial and temporal information will allow us to calculate numeric representations
(geometries, calendar periods) corresponding to the information described in the
models.
Before describing our approach, which combines models and processes drawn
from these different fields for the construction of a multicriteria search engine, we
present the notions of IR, GIR, NLP and qualitative reasoning, some of which have
already been briefly introduced.
Manning et al. [MAN 08b]defineIRintwomainprocesseswith,ontheonehand,
techniques of document indexing and retrieval mechanisms, on the other. Thus (see
section1.2),thedocumentsarerepresentedbyasetofdescriptors(terms,forexample)
organized in an index.The user'sneed (query) is dealt with in a similar way: it is also
represented by a set of descriptors. Afterwards, matching mechanisms compare the
descriptors of the query with those contained in the index in order to build a list of
relevant documents.
GIR differs from IR in its explicit recognition and modelization of space and time
in the frame of indexing processes and IR [LEI 07, ALO 08]. In a GIR system, not
only the key terms are indexed, but also the “spatial terms” with the corresponding
geo-localizations called spatial footprints and the “temporal terms” with the
corresponding intervals of time called time stamps. IR is, in this case, completed by
the comparison of the spatial footprint or the time stamp of a query with the spatial