The subject of my thesis was in the context of flexible querying of distributed information systems.
Flexible querying consists in integrating user preferences into queries, addressed to relational databases. In my case, user preferences are expressed by fuzzy bipolar conditions made of an acceptance component and an optimization component. Concerning the reasoning, it is based on fuzzy description logic of the well-proposed bipolarity. The combination of the bipolarity and the description logic is aimed at personalizing the querying process. The topic was "Personalized querying of information systems dedicated to transportation: a fuzzy bipolar approach". The manuscript (pdf in french only) is downloadable on ISITRAM-Thesis_Version_finale_PostSoutenance-TAMANI.pdf. The thesis was conducted within the Pilgrim team at IRISA under the supervision of Daniel Rocacher and Ludovic Liétard.
I was particularly interested in the personalization of querying systems taking into account both the distribution of data sources and user preferences. The latter is expressed by fuzzy bipolar conditions and modeled in the framework of fuzzy sets theory. I have proposed in my thesis:
a new formalism for user preferences based on bipolarity expressing mandatory preferences (constraints) and optional preferences (wishes),
a bipolar relational algebra and its associated SQL-like query language: bipolar SQLf,
query rewriting enrichment algorithms based on DLR-Lite ontologies,
a practical use-case of these notions in multimodal route planning.
To take into consideration the personalization when the amount of data to be considered is huge, I studied different description logics developed for querying databases. Description Logics are knowledge representation formalisms, which can be used to define ontologies and to develop Semantic Web applications that require querying large amounts of data. They are exploited in my approach to target subsets of data that are the most suitable to answer queries involving bipolar user preferences.
Furthermore, the study of data involved in public transportation systems permitted us to identify their different types, particularly spatiotemporal data. The ontology studied (OTN) allowed us to specify this kind of data type and to process them at the level of applications (route planning in our case). In addition, the implementation of a prototype is also possible since PostgreSQL BDMS handles this data type through its PostGIS component. We can also make use of the geospatial data model developed by Oracle.
I defended my thesis on April 23rd, 2012, and received my Ph.D. degree from the judging panel composed of the following professors and Lecturers:
Henri Prade, Director of research at CNRS, IRIT, Toulouse
Patrice Buche, Researcher-HDR, INRA, Montpellier
Salem Benferhat, Professor at CRIL, University of d'Artois, France (Head of the panel)
Sébastien Ferré, Lecturer at IRISA, University of Rennes 1
Ludovic Liétard, Lecturer-HDR at IRISA, University of Rennes 1
Daniel Rocacher, Professor at IRISA, University of Rennes 1