General Conclusion - Geographical Information Retrieval in Textual Corpora - page 93

In the framework of the ANR MOANO project, discussed in Chapter 1, we
contribute to the construction of a botanic ontology and to the design of a system
of indexing and retrieval of botanic information using this ontology. We reuse the
ontological and terminological resource (OTR) model and the information indexing
andretrievalengineTextVizdevelopedintheframeworkoftheANRDynamo 1 project
[REY 09]. The current work focuses on the digitalization of “plant” files extracted
fromtheClauseVillemoringuide 2 andtheirindexingbasedonabotanicontologybuilt
inasemiautomaticwayfromresourcestakenfromdevotedWebsites.Ourcontribution
aims at information indexing and the retrieval in textual corpora. Following the
previously mentioned ANR Dynamo project, we propose implementing an indexing
Webservicethat,ininput,takesadomainontologyandatextand,asoutput,produces
the same text enriched by annotations corresponding to the concepts of the ontology.
Concerning the IR, we have experimented queries focusing on flowers, their colors,
seasonal properties, diseases, etc. The results of TextViz are sometimes particularly
noisy (i.e. many non-relevant retrieved documents): Figure 4.1 presents a subset of
resultsobtainedforthequery Q7 (“Whichflowersblossominwinter?”).Theconcepts
of “flower” and “winter” have been well detected in the query and in each of the
documents. However, there is no relation of blossoming between winter and the plant
considered in the documents agapanthe0.xml and buglosse1.xml. Let us therefore
consider the following approach in order to avoid this kind of side effect:
1) Use the concepts extracted from the query to look for relations that, in the
ontology, link these concepts (e.g. floweringPeriod, plantingPeriod).
2) Select the relation expressed in the query among the propositions in this list
(e.g. floweringPeriod).
3) Searchalltheflowersintheontologythatareassociatedwiththetargetconcept
via this relation (e.g. “flower” in the relation floweringPeriod with “winter”).
4) Finally, search for the documents of the corpus dealing with the flowers
extracted from the ontology in such a way.
Figure 4.1. Example of a query submitted to the TextViz IR engine
1 http://www.irit.fr/dynamo/.
2 http://www.vilmorin-jardin.fr/vilmorin-graines-et-semences-pour-le-jardin-9-11-1-1.htm.