Finally, this final experiment has also allowed us to show that it is thus possible to
federate search engines which do not support the modal operators of requirement and
This chapter is devoted to GIR. We have designed, implemented and tested a
number of propositions contributing to the development of a multicriteria IRS,
adapted to the spatial, temporal and thematic dimensions of geographic information.
We have thus proposed a generic approach to standardization applicable to each
of the dimensions of the indexed geographic information [PAL 12b]. The PIV spatial
and PIV temporal prototypes implement spatial and temporal indexes generalized by
administrative and calendar tilings, respectively.
Moreover, we have developed a partially compensatory multicriteria IR approach
(CMRP model) supporting roles which, associated with every search criterion,
increase its expressive power [PAL 12a]. The PIV 3 prototype supports the
multicriteria IR and combines results from spatial, temporal and thematic IRSs: in
the framework of our tests, PIV 3 brings together PIV spatial , PIV temporal and
Terrier [OUN 05].
We have studied information needs formulated in the contexts of IR linked to
digital libraries and expressive queries. We propose a partially compensatory
multicriteria query model in which each criterion can be refined by a role describing
its importance in the query. Thus, the CMRP model supports the definition and the
processing of preferences and requirements in such queries. The requirements are
expressed by modal operators that emphasize the mandatory, exclusive or neutral
character of a criterion, whereas the preferences are expressed over a criterion by
other operators such as factors of promotion and demotion. The CMRP model
integrates these operators in a partially compensatory aggregation function
combining the lists of results from each of the search criteria. These modal operators
satisfy a higher need for expressiveness and allow the promotion or demotion of
documents in the final list of results. We have designed the PIV 3 GIR engine (see
Figure 3.13) that implements CMRP for purposes of validation and testing of our
propositions. This system takes into account the geographic dimensions of
information by involving domain-specific IRSs. Each domain IRS being dedicated,
respectively, to spatial, temporal or thematic information, the search engine allows
the end user to exploit these dimensions in their query. Each search criterion is
therefore part of a particular dimension and can be associated with roles.