the approaches of partially compensatory aggregation, proposed in multicriteria
IR, authorize the weighting of each criterion according to the desired level of
importance [FAR 08, DA 09, DA 12]. We propose adapting these approaches to
design a partially compensatory GIR model which we have named CMRP (“criteria,
matching operators, requirements, preferences” quadruplet for query aggregation):
each search criterion (1) corresponds to one of the three spatial, temporal or thematic
dimensions; (2) names a matching operator compatible with this dimension and,
of the criterion for the search. We study the interest of several types of IRS coupling,
including the new CMRP approach that takes on an increased power of expression.
Thus, we test our two other working hypotheses:
- 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 (offering greater power of expression to the user) of
spatial, temporal and thematic IRSs gives better results than a “classic” coupling.
We propose a new framework of evaluating GIRSs and measure the precision of a
new PIV prototype implementing respectively classic and advanced (CMRP)
couplings of spatial, temporal and thematic IRSs.
and thematic tiles in section 3.4.1. This involves the creation of indexes of a higher
level of abstraction. We will then describe, in section 3.4.2, how the vectorial model
is applied to such indexes and, finally, we propose a new aggregation model (CMRP)
of the results of spatial, temporal and thematic IRs in section 3.4.3.
3.4. Proposition for indexing by tiling and multicriteria IR in textual corpora
We have chosen to deal with each dimension of the geographic information in a
specific way. Before any combination, we have shown interest in standardizing the
representations as well as the process flows relative to these data (see section 3.3).
Our objective is therefore to create a generic approach to standardization that can be
applied to each of the geographic dimensions.
3.4.1. Standardization by tiling
We complete the spatial and temporal indexing flow described in Figure 2.6 in
section 2.4. A fourth stage, dedicated to the generalization of unitary indexes, is thus
added to the process flow (Figure 3.2).
The fourth stage, thus, applies to unitary indexes in order to create a second level
of index based on tiles. For the spatial dimension, for example, generalization