Stefano Spaccapietra

Stefano SpaccapietraStefano Spaccapietra
EPFL - IC, Station 14, INJ 236   
1015 Lausanne   
Phone:  +41 21 6935210

DVP term expires December 2014

Stefano Spaccapietra is full professor at the School of Computer and Communication Sciences, Swiss Federal Institute of Technology, in Lausanne, Switzerland, where he chairs the database laboratory. He has been an academic since 1969, when he started teaching database systems at the University of Paris VI. He moved to the university of Burgundy, Dijon, in 1983 to occupy a professor position at the Institute of Technology. He joined EPFL in 1988.

His Ph.D. is from the University of Paris VI, in 1978. His first interests were in the area of data modeling and distributed database management. Together with Christine Parent, he developed a conceptual modeling approach (based on ERC+, an extended entity-relationship data model suited for the support of complex objects) to serve as a pivotal element in a heterogeneous data architecture. Christine Parent and Stefano Spaccapietra have jointly authored many papers on the specifications of the ERC+ approach and its different aspects, and on its use in solving semantic conflicts in heterogeneous federated databases. They gave several tutorials and invited talks on the topics.

Since he joined EPFL, he had the opportunity to develop R&D activities on visual user interfaces, which eventually resulted in the SUPER prototype. This lead his laboratory to participate into two European projects (Helios and Swap). He also promoted research on semantic interoperability, including a joint research project (FEMUS) with the Polytechnic in Zürich (Prof. Schek, ETHZ). Starting 1995 a new research direction on spatio-temporal data modeling has become the major driving force of his activities. Together with Christine Parent, they turned ERC+ into MADS, a conceptual spatio-temporal data model. MADS has already been used in the design phase for several applications. It has been recently extended to include multi-representation support, thanks to a new European IST project named MurMur. Other running projects in the lab deal with user interfaces, multimedia databases (Swiss NCCR IM2 project), various modeling issues (views, schema evolution, data model translation), ontologies (European DIP Integrated project and KnowledgeWeb Network of Excellence), and location-based services for mobile users (Swiss NCCR MICS project).

Stefano Spaccapietra currently chairs the IFIP 2.6 Working Group on Databases. Previously, he chaired the Database Group of the Swiss Informatics Society (1997-2000), the ISO Working Group "DBMS Coordination", ISO/TC97/SC5/WG5 (1981-1985), and the ISO Reporter group, ISO/TC97/SC21/WG3 (1985-1986). He was also chair of the "Distributed databases" Working Group, 1975-1980, within AFCET -Informatique (French Computer Society).

Mobility Data: Modeling, Management, and Understanding
The scope of this tutorial is to present the research challenges and their up-to-date solutions to represent and analyze the movement information collected by mobile devices. The first part of the tutorial covers the data modeling issues. The focus is on showing how the traditional vision of movement as mere spatio-temporal data can be turned into a semantic vision that organizes data into meaningful trajectories whose semantics is driven by applications requirements. This change in vision, from raw data to semantic data, is the basic novel feature that is leading towards an explosion in the number of applications using mobility data. Such applications, however, require more than just the provision of trajectory data. Given the huge amount of available data, techniques for synthesizing this information have crucial importance. The second part of the tutorial therefore covers data warehousing techniques specifically devised for the management of trajectory data. The third part explores how data mining and knowledge extraction techniques complement data warehousing in creating the necessary abstractions that provide readily useful trajectory knowledge in a compact form.

Mitigated Determination of non-Significance
Le glissement de la société vers une communication interplanétaire remet à jour quelques questions fondamentales. Comment être sûr de bien comprendre les paroles de l'autre ? Au delà des mots, comment interpréter correctement la pensée de l'autre compte tenu de sa culture et des circonstances, personnelles ou environnementales, durables ou occasionnelles, qui peuvent l'influencer à un instant donné et dans un lieu donné ? Comment assurer cela sans chercher à générer un système qui sache tout de tous et de toutes choses ?

Dans ce contexte la sémantique retrouve ses lettres de noblesse et les ontologies se placent au c?ur du débat et des solutions. Autour d'une ontologie, quelques principes de base permettent au système d'atteindre un double objectif. D'une part acquérir progressivement la connaissance pertinente pour un domaine donné, comme le font les enfants en bas âge, et ainsi développer l'intelligence du fournisseur de services d'information. D'autre part respecter les différences, en adaptant l'offre du fournisseur aux besoins variables de ses utilisateurs.

Conceptual models versus ontologies
The overwhelming use of Internet emphasizes the need for efficient techniques aiming at facilitating the understanding of the information that is found over the Web. This need for understanding is currently addressed through the development of ontologies. The push towards an effective use of ontologies as a means to achieve semantic interoperability is, in our opinion, shifting the focus from purely taxonomic ontologies to more descriptive ontologies. These would namely provide agreed descriptions of the data structures representing the complex organization of objects and links of interest within the targeted domain.

This seminar will analyze the requirements for such descriptive ontologies, and contrast the requirements to the functionality provided by some current representative approaches that have been proposed for ontology management. Selected approaches originate from research in artificial intelligence, knowledge representation, and database conceptual modeling.