Sébastien Ferré - professional page
|
|
|
Address
|
: |
Irisa/Ifsic
Université de Rennes 1
Campus de Beaulieu
35042 Rennes cedex
France
|
|
|
Email
|
: |
ferre@irisa.fr |
|
|
Phone
|
: |
+33 2 99 84 75 70
|
|
|
Fax
|
: |
+33 2 99 94 71 71
|
|
|
 french
 english
 esperanto
|
Welcome
to my web page.
I am an assistant professor in computer science at the University of
Rennes 1,
and I am a member of the research team LIS in the Irisa laboratory.
Curriculum Vitae
Subject of my PhD
Title: Logical Information Systems: a
logico-contextual paradigm for querying, navigating, and learning.
Keywords : information systems, file
systems, databases (relational, object and deductive), concept
analysis, querying, navigation, knowledge representation, description
logics, views,
software engineering.
Subject : Many computer environments
are
based on rigid navigational structures: file systems, programming
environments, hypertext, Web, etc. This kind of organisation is not
well adapted to the diversity of the users and their objectives, and to
their evolution. More recent proposals define structures better adapted
to the diversity of points of view, but they are often either
overspecialized (e.g., in software engineering), or ad-hoc (e.g.,
access to files by their
content).
On the contrary, we want an information
structure that
- is not too specialized in order to use it
in various domains (e.g., software engineering, digital libraries, or
information servers for a large public),
- takes into account informations that are
not only intrinsic to represented objects (e.g., circumstances of the
entry of an object in the information system),
- can be implemented at a low level of the
system architecture (e.g., level of the file system) in order to make
it accessible from all applications.
We propose to study navigational structures based
on properties of terminal objects. They have to be navigational to make
it possible to steadily explore a set of objects, but the links
supporting the navigation will not be induced from the representation
of objects. The contents of terminal objects will contribute to the
elaboration of intrinsic properties of these objects, but some
extrinsic properties will also be elaborated, especially at the
creation of these objects.
This project is composed of a fundamental aspect,
which has to offer a formal modelisation of these structures, and an
applied aspect, whose aim is to realise a prototype of file
system so as to validate chosen options. The first formalisms to
consider
are on one hand Concept Analysis for modelling navigational structures,
and on the other part description logics for representing properties
of the objetcs. Preferred applications will be software engineering
and information servers for a large public (e.g., catalogues, guides).
Summary: Both main paradigms of information retrieval (IR),
which are navigation and querying, are often disconnected in existing
systems. Hierarchical systems (e.g., file systems, Web) have a rigid
navigation structure that is not appropriate to all users and all uses;
what is compensated by searching tools (e.g., command "find", search
engines). These tools, based on querying, are more flexible but are
difficult
to use for non-expert users and make it hard to control the number of
answers. Therefore, it looks like necessary to tightly combine
navigation
and querying in IR.
In order to achieve this combination, we use as a fundation Concept
Analysis (CA). Given a description of the objects of a system, it
enables to automatically build a navigation structure called "concept
lattice", where concepts plays as both directories and queries. As
object descriptions are limited to sets of attributes in CA, we first
generalized CA by replacing them by formulas of an almost arbitrary
logic. We feel this genericity is important to handle various
applications. Logical Information Systems (LIS) are therefore defined
by the combination navigation/querying, the use of logic (for
descriptions, queries, navigation links), and logical genericity.
On this basis, we developed several mechanisms intended to
facilitate
expression and discovery of knowledge. Some knowledge about an
application
domain can be expressed through a terminology. A human-machine dialog,
based
on the concept lattice, allows the user to retrieve objects
(navigation),
and discover patterns among objects (knowledge discovery). A learning
mechanism
helps the user to classify new objects. To end, a prototype has been
developed, allowing experimentation of these mechanisms through
concrete applications (e.g., bibliographical references, emails,
receipts). It is generic in the sense that it does not depend on the
logic used in each application. To help the building of these logics,
we realized a toolbox of logical components, which can be freely
composed.
Teaching Activities
- Introduction to ILP: lecture to undergraduate
students (UWA), spring 2004 (1h) [PowerPoint]
- Final Year Placement: supervision of Cyrille Dufils,
engineer
student at INSA de Lyon, fall 2003 (4 months)
- BSc Final Year Project: supervision of 4 students
(University of Wales, Aberystwyth), 2003-04 (1h/week)
- Prolog: practicals for second year BSc students
(University of Wales, Aberystwyth), fall 2003 (12h)
- Java: practicals for engineer students (DIIC2 and DIIC3, IFSIC), fall 2001
(24h)
- UNIX: practicals for graduate students (DESS CCI,
IFSIC), fall 2001 (8h)
- Functional Programming (with Objective Caml):
practicals for engineer students (INFO3, INSA), spring 2001 (28h)
- lexical, syntactic, and semantic analysis: practicals
of for engineer students (INFO4, INSA), fall 2000 (72h)
- Functionnal programming (with Caml Light): practicals
of for engineer students (INFO3, INSA), spring 2000 (28h)
- Context-free languages: seminars for engineer students
(INFO3, INSA), fall 1999 (16h)
Links
Last update: 06/03/2007.
