Multicriteria Information Retrieval in Textual Corpora - Geographical Information Retrieval in Textual Corpora - page 86

3.6.2. Perspectives
In the framework of the CMRP model and its implementation in PIV 3 , the
number and power of expression of the IR operators correspond to the proposition of
a dynamically customizable formula of result aggregation. This approach makes the
task of the end users more difficult concerning the elaboration of queries as well as
the control of their range. We could study two categories of research interfaces. The
first category, dedicated to regular users, would enable them to choose between
predefined roles in order to add expressiveness to the search criteria: see Figure 3.13
with customizable requirement and preference criteria. The second, dedicated to
advanced users, would allow them to customize their queries by specifying modal
(ˆ, +, −) and matching operators (=, ⊆, ∩).
Concerning the MIDR_2010 test collection, a greater corpus and set of topics
have to be established in order to conduct a larger scale experiment. However, this
evaluation provides several interesting first results and can simulate the use of
unexploited operators in GIR and IR systems (e.g. Lemur [OGI 01] or
Lucene [GOS 05]). Furthermore, we can note that the completeness of the proposed
language is not assured, even though CMRP supports the “and” and “or” operators
through those of requirement and preference.
Finally, let us note that while the principle of multicriteria IR is well suited
to geographic information, as we have shown by conducting an experiment on the
TREC-8 collection, it could equally be applied to the combination of different textual
IRSs and/or other specialized ones such as opinion IRSs [ROC 09, LI 10], for
example.