I'm Research Scientist at Xerox Research Centre Europe (XRCE) since 2006, Grenoble, in the Parsing & Semantics area (former manager: F. Segond, new manager: TBC). My current research topics are around semantic information extraction from text, with a focus on the following approaches:
SynC3 is a European project involving 9 partners (research labs, journalistic organizations, companies such as Xerox and Google). (official web site).
Overview: SynC3 addresses the business areas of news reporting and public opinion formation to deliver a framework that combines news events and their relations with users and their opinions and produces the SynC3 graph of the commentaries of blog users on current news. Towards this direction, SynC3 will develop accurate methods for connecting users to other users and news content.
XRCE is involved in this project as a technology provider. Namely to provide linguistic information from news articles and to cluster the news articles into news events based on an approach combining statistical processing and linguistic methods applied to both low level and high level features.
In this project, I'm strongly involved in the linguistic information extraction part and partly involved in the news article clustering part.
Infom@gic is a french project from the "pôle de compétitivité Cap Digital" involving 30 partners (research labs or companies such as Xerox, EADS or Thales). (official web site).
Overview: develop new tools for information extraction, retrieval and analysis from multi-source data (text, image, sound, structured data).
XRCE is involved in this project as the leader of the text processing and in different tasks such as information extraction using linguistic methods, different use-cases such as risk detection task. XRCE was also involved in the UIMA platform development which was used during the project.
I have been strongly involved in these aforementioned tasks.
This is the 4th International Workshop on Semantic Evaluations. We (Brun et al., 2007) participated in the "named entity metonymy resolution" task and our system came second and third respectively for the location and organization named entity types.
Project involving the LATTICE (Paris), CRISCO (Caen) and ERSS (Toulouse) laboratories.
Overview: theoretical study of verbal polysemy and its relation with syntactic constructions. Implementation of this theoretical study based on a graph approach.
I was responsible for the integration of all the modules used for this project: Integrating the parsers "Syntex" (D. Bourigault) and "Wims" (E. Giguet) with the graph processing software "Visusyn", importing text data from "Frantext" (ATILF) and implementing the sense calculation model.
2005: Ph.D. in Cognitive Sciences (specialized in
Computational Linguistics)
Title: A continuous model for Verb Sense Disambiguation (Polysémie verbale et
calcul du sens).
supervisor B. Victorri, Lattice lab (CNRS-ENS).
2002: Master (Graduate Degree) in Cognitive Sciences (specialized in Computational Linguistics)
2001: M.S.T. (equivalent B.A.) in Mathematics and Computer Science applied to Social Sciences.
2009-today: Board's member of the UFR "Sciences du Langage" (Linguistics), Grenoble 3.
2009-today: Board's member of ATALA (French National association on NLP).
2007-today: Member of the jury for XRCE patent proposal evaluation.
Program committee member/reviewer: EACL 2009, Semeval 2007, LTC 2007, FinTAL 2006, TALN 2008-2012 and TAL journal.
Systems: Windows, Unix
Programming Languages : Matlab, Java, Python, Perl
Frameworks : UIMA, Hadoop
Languages : French and English