| Network
of Excellence |
CARiMan |
The
Internet Portal for Computer Aided Risk Management |
MEMODATA,
founded in 1989; is a SME settled at Caen (Normandy). M. Dominique Dutoit is its
director. The activity of the company is Language processing. Language
processing has main purposes : make the human language accessible to the
computers, make the computers more accessible to human beings. To perform this
task, the company is following several main lines.
Lexicology
and terminology
Several
alphabetic dictionaries have been written by MEMODATA, for electronic publishing
purposes (for Goto, Lexibook). Moreover, MEMODATA has a dictionary for
translation (5 languages linked through the semantic dictionary of the company).
But MEMODATA has also published under its name, DICOLOGIQUETM,
a user-friendly semantic dictionary (downloadable on its Web site http://www.memodata.com).
Phonetics
The
access to proper nouns databases (nitcheà
niestzche), phonetic spell checking, rhymes (…) are properties of the
phonetics units designed and sold by the company (Universalis, Hachette, Xerox,
Sligos, Unisys, Thomson, Cap Gemini Telecom…) for French and English.
Grammar,
conjugation, inflexion, derivation, synonym.
For
each of these fields, MEMODATA can sell a set of linguistic resources, without
royalties such as a database featuring the words with their inflected model (ex
: avoir à
j'ai, tu as, avions à
avoir ou avion). These products are packages in an electronic form (references :
De Boeck (Grévisse), Lockace, NTT, ELRA…).
Artificial
Intelligence and semantic nets.
It's
the main core of the works of the company since the origin. The "semantic
net" is called "The Integral DictionaryTM ".
The component of The Integral DictionaryTM allows the semantic
interpretation of a text thanks to its millions conditional analogies, its
current knowledge and its generation of paraphrases. An example is given by the reverse
dictionary, which gives rhabdomancian for an answer as person who
predicts the future in looking at the springs. Another example is given in
the semantic summary of texts.
These
tools are used in the industry and for R&D, under the name of SemiographTM.
The Semiograph is a set of API (in Java) which uses the data of The Integral
DictionaryTM. One of the main purpose of the Semiograph is to build a
new kind of search engine for the Internet or search agents. The main
characteristic of this project is to forget the words of the texts and to
remember the deep knowledge by assessing the meaning.
The
team of MEMODATA has participated in several national or European research
projects: Lockace (MENRT, search Engine), STOULS (MENRT, intelligent search
agents), CRISTAL (LRE, DGXIII FP3), MARLEN (LEONARDO, help for creative
writing), EUROWORDNET (TELEMATICS, LE, DGXIII, FP4), AGIR (MENRT, semantic
signature of texts), VIVIAN (ITEA, search agent for PDA), EuroWordNet (DGXIII),
IVOMOB (MENRT), BALKANET (IST, FP 5).
MEMODATA
has developed direct collaborations with private research centers (XEROX
Research center, ALCATEL CIT, Thomson, Banque Paribas…). The works of
Dominique DUTOIT in the company have also given a University thesis : "Some
sense àtext
and textàsense
calculation using an aprioristic universal semantics."
Integrate
public knowledge base
It
exists number of public knowledge base and this number should increase in the
future, as the open source code in the software domain. One important work is to
develop an abstract model that can integrate and use these databases.
KEY
PERSON CV9
Dominique
Dutoit (40) has a Ph.D. in the area of computational science and linguistics.
Manager of the company MEMODATA since the beginning (1989), he has managed
several research programs in accordance to the goal of the company (Natural
Language Processing, semantics). In the same area, Dr Dutoit teaches
computational linguistics in the University of Caen. Some recent papers
published in famous conference (ECAI, ICCSA, CoLing) are some results of the
academic work.
Topics
interest
A1) Communication
patient - physician
Help
to access to common vocabulary
A3) communication
CAMS-CAMS
Study
of the strategies tested by Patimon. Relative efficaciousness of standards (RMI,
CORBA, XML, SOAP …) to exchange knowledge
Have
a look on multi-agent systems in comparison to the concrete goal of this
communication.
S1) communication
Physician -CAMS
Simplify
the access
S2) KM & DBS
Provide
a strong link between specialized thesaurus such as Medlin and the common
language. Integrate these thesauri into a multilingual DBS.
D2
– D3 – D4 and (?) M1, M2
Recent
Publications
[MEMO1]
Dutoit D., Poibeau T. : « Inferring knowledge from a large semantic
network », Août 2002, full paper, acte de Conference on Computational
linguistics, COLING TAIWAN
[MEMO2]
Dutoit D., Poibeau T. : « Generating extraction patterns from a large
semantic network and an untagged corpora », Août 2002, acte de Workshop,
COLING, TAIWAN.
[MEMO3]
Dutoit D, Nugues P. : « A lexical network and an algorithm to find
words from definitions », acte de European Conference on Artificial
Intelligence, ECAI 2002, LYON.
[MEMO4]
Dutoit D, Poibeau T. : « Évaluer l’acquisition semi-automatique de
classes sémantiques », acte de TALN 2002.
[MEMO5]
Dutoit D., Poibeau T. : « Evaluating resource acquisition tools for
information extraction », May 2002, full paper, acte de Language
resource and evaluation, LREC, Las Palmas
[MEMO6]
Dutoit D, Nugues P. : « The right word », May 2002, full paper,
acte de Language resource and evaluation, LREC, Las Palmas
[MEMO7]
Dutoit D, Nugues P., P. de Torcy: « The Integral Dictionary : a
lexical network based on computational semantics », Springer Ed., May
2003, ICCSA International Conference on Computational Science and its
Applications, Calgary, Canada
| Network
of Excellence |
CARiMan |
The
Internet Portal for Computer Aided Risk Management |