Paris Institute of Technology (Paris Tech)

Ecole Nationale Supérieure des Télécommunications (ENST)

LTCI URA820 \ Sciences et Technologies de l'Information et de la Communication \ Centre National de la Recherche Scientifique (STIC CNRS)

Networks and Computer Science Department (INFRES)

Information-Interaction-Intelligence research group (3I)

Information processing is facing incrementing knowledge and information volumes. This requires smart systems and efficient, user-friendly human interfaces. In response to this challenge, the I3 group research priorities are :
Since its creation in 1998, our group has not only performed advanced research in the three scientific domains it represented at the begining - artificial intelligence (AI), information visualisation (IV) and data bases (DB) - but experimented multi-disciplinary approach on more general issues as knowledge management, distributed intelligence, learning strategies, evolution of competences in communication... particularly since it has widened its scope to linguistic and philosophical research.

Our group is or has been recently involved in research projects such as ITEA-Ambience (adaptative environmement, dynamic networking, distributed intelligence for mobile persons), GET-"campus mobile" (e-learning, virtual worlds, navigation through large information spaces), GET-"e-parcours" (semantic web, e-learning, community web) ...

Permanent members

Formal methods in semantics generally refer (i) to the use of logical tools for the representation of the meaning of natural language expressions and (ii) to the postulate that natural language meaning can be represented in terms of truth values by a compositional calculus. This amounts to apprehending the meaning of a sentence or a discourse from he meaning of its parts and to evaluating it with respect to a given world. A new community of formal semanticists is today appearing in France. The GDR 25-21 witnesses this growth and, for this CNRS-financed research group, I am actually organizing the first international colloquium at the ENST in collaboration with Paris IV and the Institut Jean Nicod. 

 

I have reoriented my research in 2002 and today, my work in formal semantics specifically brings on complex events. The question I address concerns the existence of events – and of complex events in particular - as ontologically existing entities. This question has to be answered with respect to a formal theory of meaning and in the davidsonian philosophical framework. Davidson has first posited the existence of events as individuals, on the same level as the Aristotelian quality and substance. His hypothesis needs a confirmation on an empirical level and still needs to receive a formal evaluation. I am also researching how the complexity of events is reflected in language and in cognition with respect to particular syntactic and semantic constructions such as subordination, plurality and comitativity. I am also investigating how this complexity is acquired and represented. Formal methods give a computational account of my hypotheses. I plan to pursue this research for at least a couple of years, and to publish a second book on this issue.


During the past two years, in this scientific framework, I have been working on French prepositions, and polysemy in particular. I have been addressing the wittgenstenian question of knowing how and when to apply a word with many senses in a particular context. I treated this subject in my PhD - obtained in 2000- and have been pursuing it in my 2001 research and publications. A book on this issue will appear soon at the L’Harmattan Publisher: “Principes d’identification et de categorisation du sens”.

Sabbatical recent (2002) visitors

Postdoctoral or temporary members

PhD students (recently or nearly to be performed dissertations)

Any query expressible on a set of records can be expressed on an array of records. The reverse doesn't hold; the notion of order is missing. Queries like ranks (select the first n), n-tiles (select all in the top n-tile), deltas (select last - first in a group), or vector aggregates (moving average) all rely on some ordering on input or on resulting data. And Finance, Biology, and Linguistics domains, for instance, have long taken advantage of such a possibility. Manipulating data as sequences or arrays, as opposed to sets, is a critical new feature if we are to incorporate order into Database Systems.
Previous attempts were successful in showing that sequences do "fit" in a relational context. They can be manipulated by extended versions of relational algebra. They can be queried with extensions of SQL. Nevertheless, the approaches taken to do that had led to queries that are somehow difficult to express and thus hard to optimize.
The notion of order is intrinsic to AQuery's data model and query language. Data is represented by "array tables," the arrables, which capture order relationship. Every algebra operator has its impact on order, or absence thereof, clearly defined. The query language is as faithful as possible to SQL, differing only in its array-flavor. As a result, AQuery's optimizer often finds liner-time plans for queries that would otherwise take more complex solutions.



last update on 2003/02/10 by  saglio@enst.fr