Geoscience Reference
In-Depth Information
DSSs have to conform to the preferences of the user. A preliminary system
adjustment phase is therefore necessary. This setting phase is dependent on the user:
for each criterion, the user is encouraged to classify the possible results in
categories/levels of interest and also to define the difference felt between each pair of
categories.
3.3.1.4. Spatial grouping and GIR
Lieberman et al. [LIE 07] and Martins et al. [MAR 08b] tackle the problem of
combining the place references given in a text in order to find the encompassing
geographic scope that the document discusses as a whole. They propose the
standardization of spatial information by mainly searching for the “focus” of the
document, in other words the main area which it relates to. The advantage of this is
the simplification of the spatial search, given that a single piece of spatial information
is associated with each document. The administrative divisions being hierarchical
(the world is divided into continents, followed by countries, regions, cities, etc.) and
meaningful, Lieberman et al. [LIE 07] and Martins et al. [MAR 08b] implement this
type of standardization. It is thus possible to regroup several cities within the same
region, for example.
By applying a similar approach to the temporal and thematic dimensions, in other
words by obtaining the spatial, temporal and thematic scopes of the document, the
combinations would apply in a logical way. Indeed, each criterion subject to a
generalization-type standardization, the criteria would then be consistent.
Nevertheless, as Lieberman et al. [LIE 07] explain, unless the inaccuracy of the
results is very strongly increased, it is very difficult to obtain a single focus if the
document contains very distant spatial information. Thus, for example, for a
document evoking the city of Biarritz (in Southwestern France) and the city of Lille
(in the North of France), the process of generalization will propose France. This form
of grouping can, therefore, lead to an excessive generalization [LIE 07].
3.3.2. Combination of criteria
We have studied the approaches dedicated to the standardization of criteria in
fields such as IR, multimedia IR, DSSs and GIR. We will now show interest in the
approaches to the combination of criteria in these same fields. Nevertheless, it has to
be specified that, depending on the field, the term used is not combination because
even if the aim is to unite several criteria, the context, the number of criteria and the
methods can be very different. It is called criteria fusion for multimedia IR and
criteria aggregation for DSSs. Criteria fusion, on the one hand, allows us to gather
sets of information extracted from different documents (e.g. images and texts) or
from different parts of the same document (e.g. a video can be broken down into a set
Search WWH ::
Custom Search