Multicriteria Information Retrieval in Textual Corpora - Geographical Information Retrieval in Textual Corpora - page 72

resultsofournewspatial(PIV spatial prototype)andtemporal(PIV temporal prototype)
IRapproaches.Thesewillalsoallowustodeterminewhichtilingandwhichweighting
formula give the best results for the spatial dimension and the temporal dimension.
We now have homogeneous spatial and temporal IRSs, because they are based on
the same model of data representation and on the same IR model. In the following
chapter, we will focus on the design and implementation of a GIRS combining the
spatial, temporal and thematic dimensions.
3.4.3. Multicriteria IR applied to tiling: PIV 3
Indigitallibraries,IRpresentsanumberofspecificities.Thecontentofthecorpora
generally references specific fields (fauna, flora, art, sport, etc.). Consequently, the
needs as well as the vocabulary of the users are more precise. Moreover, the users
of these repositories are familiar with the manipulation of the search operators. Thus,
an advanced user interested in documents dealing with “Pyrénées mountains but not
thoseofGavarnie,inthe19 th Century,ifpossible,unrelatedtoascents”[PAL 10a]has
to be able to express this type of need and navigate in the set of document fragments
retrieved by the IR system. The more expressive the query is, the higher the number
of associated criteria and operators will be. This must finally be completed by the
particularityofgeographicinformationthatweapproachviathespatial, temporaland
thematic dimensions. We locate these propositions in the line of the study of Ray R.
Larson [LAR 09] who shows the efficiency of GIR methods in the access to digital
libraries.
During the process of need formulation, this category of users introduces more
expressiveness with the notion of role such as “should”, “has to”, “must not” to give
importanceorprioritytoacriterionortouseacriterionasanexclusionkey.Theusual
search engines (such as Google, Terrier or Lucene) propose two ways of formulating
needs:
- regular: theusergivesoneormorekeywords,thesystemcombinesthedifferent
sets of results without the intervention of the user;
- advanced: the user gives several keywords and specifies those which are
mandatory or to be excluded and can assign a weight to others.
GIRSsproposedifferenttypesofformulationsimilartotheregulartypementioned
above.Theuserdoesnotinterveneintheaggregationoftheresults;theexpressiveness
of the queries is relatively weak. Consequently, we propose an IR approach which is:
- multicriteria: the query contains several search criteria to match documents;
- geographic: eachcriterionconcernsthespatial, temporalorthematic dimension
of the document repository;