Access by Geographic Content to Textual Corpora: What Orientations? - Geographical Information Retrieval in Textual Corpora - page 14

information contained in textual documents. In the frame of the GEOTOPIA project
(CNRS TGE-ADONIS Program - Innovative tools of digital processing for the
valuation and diffusion of data), we adapt and encapsulate these processes in Web
services dedicated to spatial and temporal-annotations of documents in a Web
platform of sharing and collaborative enrichment of archival data. The GEONTO
project (ANR-07-MDCO-005-01 Project - Flow of Data and Knowledge) leads us to
use the “public” document repository digitized by MIDR for the purposes of
enriching a geographic ontology. This geographic ontology, resulting from the
processing of numerous resources, is then integrated into the spatial process flow for
a better typing of the toponyms and consequently a more adapted approach of
disambiguation during the search for representations in gazetteer-like geographic
resources (dictionary of place names, describing their physical characteristics -
nature or geo-localization of places - as well as social and economical). The
GEOCIME project (64 - General Council Project) targets the development of an
educational application allowing children and their educators to study on collections
of multimedia documents before, during and after a field trip. It is based on indexing
tools resulting from these different projects. The CIDRI project (Pau Pyrénées
agglomeration Community Project) is the extension of the PIV project. It aims to
generalize spatial and temporal indexes in order to develop an IR prototype
combining spatial, temporal and thematic criteria (limited to terms). The MOANO
project (ANR-2010-CORD-024-01 Project - Content and Interactions), in its GIR
aspect, targets the construction and integration of a domain-specific ontology
dedicated to botany. This ontology serves as a core model in the generalization of
indexes of the thematic process flow. The thematic point of view on the document
corpus used is here limited to botany. Finally, the IR engine prototypes PIV 1 , PIV 2
and PIV 3 result from these different pieces of study. PIV 1 proposes a spatial IR
engine and a temporal IR engine based, respectively, on fine spatial footprint and
time stamp representations. PIV 2 proposes a spatial IR engine and a temporal IR
engine based, respectively, on generalized representations corresponding to spatial
and temporal tilings. Thus, PIV 2 implements the vectorial IR model in both
prototypes. PIV 3 proposes a multicriteria IR engine using spatial PIV 2 , temporal
PIV 2 and Terrier IRS [OUN 05]. PIV 3 then aggregates the results coming from these
different IRSs.
All these applications are also motivations for our approach. We can thus say, by
recalling our research objectives mentioned earlier, that:
- our models support the symbolic and numeric representation of spatial and
temporal data, as well as our spatio-temporal processes of extraction, interpretation
andindexingareputtothetestintheframeoftheGEOSEM2,GEOTOPIA,GEONTO
and PIV projects;
- our spatial and temporal IR models are put in practice in the frame of the PIV,
GEOCIME and PIV 1 projects;