Geoscience Reference
In-Depth Information
NEs in a GIR system. We can mention Alexandria Digital Library (ADL)
20
,
GeoNames
21
, Getty Thesaurus of Geographic Names (TGN)
22
, GEOnet Names
Server (GNS)
23
, Geographic Names Information System (GNIS)
24
, Géoportail
IGN
25
(BD TOPO and BD CARTO of the French “Institut Géographique National”).
2.3.8. Summary
In section 2.3, we have presented the different pieces of work relative to the
content analysis of textual documents. It has allowed us to identify the elements of
the literature dealing with our issue, which is to propose models and methods to build
a spatial and temporal IR system.
A result from this study is that our work lies within the field of specialized IR.
Indeed, we wish to process only a part (spatial and temporal) of the information
contained within texts instead of working with all the terms present in the corpus.
Thus, we have to perform a targeted search to recognize the significant elements of
the text. It is necessary to carry out an interpretation of such elements; indeed,
indexing “south of Pau” without knowing the spatial significance of this phrase is not
sufficient. The linguistic approach, using the tools of NLP, seems to be the most
adapted method to perform this active retrieval. This allows us to extract specific
phrases and to interpret the semantics of the relations between the terms of each
phrase.
However, the data recognized in such a way have to be organized in indexes. It is
therefore important to interpret the spatial and temporal content of the textual
documents in such a way as to build adequate structures. This is the reason why the
work relative to qualitative reasoning allows us to interpret and structure the
information initially extracted using linguistic tools. In particular, the concept of
target/site seems to be a good starting point for the interpretation and structuring of
spatial and temporal information.
In the following section, we propose an ad hoc model of representation and
retrieval of spatial and temporal information, but inspired from the existing model, in
order to take into account the spatial and temporal relations present in texts. The
same applies to the evaluation of our propositions. The testing protocol is similar to
the existing protocol, the experimental corpus is in French and the resources cover
the toponymy relative to the Pyrenean region.
20 http://www.alexandria.ucsb.edu/gazetteer/.
21 http://www.geonames.org/.
22 http://www.getty.edu/research/tools/vocabularies/.
23 http://earth-info.nga.mil/gns/html/.
24 http://nhd.usgs.gov/gnis.html/.
25 http://www.geoportail.fr/.
Search WWH ::
Custom Search