Geoscience Reference
In-Depth Information
- The generalization of data representation is adapted to the spatial and
temporal information.
- The generalization associated with a well-tried IR model does not imply a
loss of quality with respect to the initial dedicated IRSs.
- A “classic” coupling (an arithmetic mean, for example) of spatial, temporal
and thematic IRSs gives better results than a single thematic IRS or a two-by-two
pairing of spatial, temporal and thematic IRSs.
- An“advanced”coupling(offeringgreaterpowerofexpressiontotheuser)of
spatial, temporal and thematic IRSs gives better results than a “classic” coupling.
3) Objectives: Theresearchobjectivescorrespondingtotheformulatedchallenges
mainly target the design and implementation of:
- models supporting the symbolic and numeric representation of spatial and
temporal data;
- process flows of spatial and temporal information extraction, interpretation
and indexing;
- spatial and temporal IR models;
- a model of generalization of the representation of indexed data;
- a model of multicriteria IR.
1.4.2. Approach
Our approach is divided into two main steps. First (Figure 1.4), under the
direction of Gaio [GAI 08], we have implemented process flows dedicated to the
automatic recognition of spatial and temporal named entities. We consider that,
regardless of the territory, it is possible to clutch to a single or a set of named entities
as characteristic elements of the content of a corpus of documents. The processes,
successively lexical, morpho-syntactical and semantic consist, respectively, of
extracting, validating and interpreting spatial and temporal named entities for the
purposes of annotation and/or indexing. We propose a spatial and a temporal model
dedicated to the needs of marking and representing the processed information in each
of these steps. We obtain, in particular, spatial and temporal indexes in compliance
with these models.
Texts annotated in such a way mainly target experts: librarians or archivists, for
example, who need these results in order to create descriptive instructions (metadata)
in a semi-automatic way. In addition, the indexes represent the content of document
repositoriesandsupportIRscenariosaswellasmoresophisticatedapproachesforthe
Search WWH ::




Custom Search