Geoscience Reference
In-Depth Information
1.4.1. Challenges, hypotheses and research objectives
1) Challenges: Aswementionedpreviously,theratioofqueriesintegratingspatial
criteria, for example, varies between 12.7 and 18.6% according to Excite [SAN 04],
AOL [GAN 08] and Yahoo [JON 08]. Although well tried today, the classic
approaches to IR are limited in the case of geographic search criteria [LIE 09]. It
remainstruethatwhatinterestshumansmostoftenisthetheme.However,takinginto
accountthethematicdimensionandmoreimportantlythesemanticaspectsitholdsisa
verydifficulttask.Indeed,thecurrenttoolsofIRareefficientbutlimitedtoterms.Our
first objective, therefore, is to target the spatial and temporal dimensions as privileged
entrypointsinthetexts.Theaimis,thus,tocompleteclassicIRSsbyspecificservices
dedicated to the spatial and temporal aspects. The first challenge is therefore: What
models of representation and retrieval of spatial and temporal information should be
proposed for the access by geographic content to textual corpora?
It is then a question of combining a classic IRS with spatial and temporal IRSs.
The heterogeneity of the models of representation and those of the corresponding IR
does not allow us to directly consider the combination of such systems. The second
challenge is thus: Which core model of representation and retrieval of information
should be proposed in order to prepare for the combination?
Finally, the combination has to be based on aggregation operators adapted to the
geographic context. Let us not forget the need for a power of expression highlighted
earlier. This need is partially satisfied by the operators proper to each dimension.
Nonetheless, it has to be completed with a finer formulation of each criterion: is it
mandatory, is it associated with a level of preference, is it a rejection criterion? The
thirdchallengeisthus: What advanced aggregation operators should be proposed and
implemented for a multicriteria IR combining many IRSs?
2) Hypotheses: We issue a study hypothesis for each of the challenges evoked.
Concerning the implementation of specific process flows dedicated to spatial and
temporal information, our study hypotheses are the following:
- AdedicatedspatialIRSgivesbetterresultsthanaclassicthematicIRSforIR
composed of only spatial criteria.
- A dedicated temporal IRS gives better results than a classic thematic IRS for
IR composed of only temporal criteria.
- A“rough”couplingofspatial,temporalandthematicIRSsgivesbetterresults
than a classic thematic IRS for multicriteria IR despite the numerous possible biases
linked, for example, to the heterogeneity of value domains with manipulated scores.
Finally, the implementation of a multicriteria IRS combining the results from
spatial, temporal and thematic IRSs arises the following study hypotheses:
Search WWH ::
Custom Search