Phd Thesis

• Extraction de connaissances d'adaptation en raisonnement à partir de cas
  résumé (Fr) | abstract (En) | paper | slides | bibtex

  @phdthesis{phdThesisBADRA,
  	title = { {E}xtraction de connaissances d'adaptation en raisonnement {\`a} partir de cas},
  	author = {{B}adra, {F}adi},
  	school = {{U}niversit{\'e} {H}enri {P}oincar{\'e} - {N}ancy {I}},
  	day = {20},
  	month = {11},
  	year = {2009},
  }
  			

  Cette thèse se situe à l'intersection de trois domaines de recherche : le raisonnement à partir de cas, l'extraction de connaissances et la représentation des connaissances. Raisonner à partir de cas consiste à résoudre un nouveau problème en utilisant un ensemble de problèmes déjà résolus, appelés cas. Dans cette thèse, un langage de représentation des variations entre cas est introduit. Nous montrons comment ce langage peut être utilisé pour représenter les connaissances d'adaptation et pour modééliser la phase d'adaptation en raisonnement à partir de cas. Ce langage est ensuite appliqué à la tâche d'apprentissage de connaissances d'adaptation. Un processus d'extraction de connaissances, appelé CabamakA, est mis au point. Ce processus permet d'apprendre des connaissances d'adaptation par généralisation à partir d'une représentation des variations entre cas. Une discussion est ensuite menée sur les conditions d'opérationnalisation de CabamakA au sein d'un processus d'acquisition de connaissances. L'étude aboutit à la proposition d'un nouveau type d'approche pour l'acquisition de connaissances d'adaptation dans lequel le processus d'extraction de connaissances est déclenché de manière opportuniste au cours d'une session particulière de résolution de probl&egrav;mes. Les différents concepts introduits dans la thèse sont illustrés dans le domaine culinaire à travers leur application au système de raisonnement à partir de cas Taaable, qui constitue le contexte applicatif de l'étude.

  This thesis presents some contributions in three research domains : case-based reasoning, knowledge discovery and knowledge representation. Case-based reasoning consists in solving new problems by reusing a set of previous problem-solving experiences, called cases. In this thesis, a language is introduced to represent variations between cases. We first show how this language can be used to represent adaptation knowledge and to model the adaptation phase in case-based reasoning. This language is then applied to the task of adaptation knowledge learning. A knowledge discovery process, called CabamakA, is proposed, that learns adaptation knowledge by generalization from a representation of variations between cases. A discussion follows on how to make this knowledge discovery process operational in a knowledge acquisition process. The discussion leads to the proposition of a new approach for adaptation knowledge acquisition, in which the knowledge discovery process is triggered in an opportunistic manner at problem-solving time. The concepts introduced in the thesis are illustrated in the cooking domain through their application in the case-based reasoning system Taaable, that constitutes the application domain of the study.

International journals, conferences and workshops

• A Semantic Web Representation of Product Range Specification based on Constraint Satisfaction Problem in the Automotive Industry. F. Badra, F.-P. Servant and A. Passant (2011)
  ESWC Workshop on Ontology and Semantic Web for Manufacturing (OSEMA'11), Heraklion, Greece, 29 may, 2011 .
  abstract | paper | bibtex

  @inproceedings{osema11-badra, 
  title={A Semantic Web Representation of Product Range Specification based on Constraint Satisfaction Problem in the Automotive Industry},  
  author={F. Badra and F.-P. Servant and A. Passant}, 
  booktitle={Proceedings of the ESWC Workshop on Ontology and Semantic Web for Manufacturing (OSEMA'11), Heraklion, Greece, 29 may, 2011.}, 
  year={2011},
  }			
  			

  Product Range Specification (PRS) in the automotive world is one of the most complex PRS that exists in industrial contexts. PRS plays therefore a key role in the information system of an automaker: related data pervades many systems, and numerous applications are using it. This is the case at Renault, where PRS is modelled as a Constraint Satisfaction Problem. In this paper, we study how to represent the objects, concepts and services related to such a PRS using Semantic Web standards. Plugging them into a Linked Data based architecture enables with new ways to access corresponding data and tools in the whole car manufacturing and selling process.

• Mining Sequential Patterns from MODIS time series for cultivated area mapping. Y. Pitarch, E. Vintrou, F. Badra, A. Bégué and M. Teisseire (2011)
  The 14th AGILE International Conference on Geographic Information Science. 18-21 April 2011 - Utrecht, The Netherlands .
  abstract | paper | bibtex

  @inproceedings{agile2011-badra, 
  title={Mining Sequential Patterns from MODIS time series for cultivated area mapping},  
  author={Y. Pitarch and E. Vintrou and F. Badra and A. B\'egu\'e and M. Teisseire}, 
  booktitle={The 14th AGILE International Conference on Geographic Information Science.
  			18-21 April 2011 - Utrecht, The Netherlands}, 
  year={2011},
  }			
  			

  To predict and respond to famine and other forms of food insecurity, different early warning systems are using remote analyses of crop condition and agricultural production, using satellite-based infor- mation. To improve these predictions, a reliable estimation of the cul- tivated area at national scale must be carried out. In this study, we developed a datamining methodology for extracting cultivated domain patterns based on their temporal behavior as captured in time-series of moderate resolution remote sensing MODIS images.

• Fouille de données complexes : des relevés terrain aux données satellitaires pour la cartographie des paysages agricoles. F. Badra, E. Vintrou, A. Bégué and M. Teisseire (2010)
  Conférence Internationale de Géomatique et Analyse Spatiale SAGEO'10, Toulouse 17-19 novembre 2010 .
  abstract | paper | bibtex

  @inproceedings{sageo2010-badra, 
  title={Fouille de donn\'ees complexes : des relev\'es terrain aux donn\'ees satellitaires pour la cartographie des paysages agricoles},  
  author={F. Badra and E. Vintrou and A. B\'egu\'e and M. Teisseire}, 
  booktitle={Conf\'erence Internationale de G\'eomatique et Analyse Spatiale SAGEO'10, Toulouse 17-19 novembre 2010}, 
  year={2010},
  }			
  			

  Dans cet article, nous présentons une approche préliminaire de caractérisation des paysages ruraux et de leurs systèmes de culture à partir de techniques de fouille de données (recherche d'itemsets fréquents). Cette méthode permet de coupler des données de relevé terrain aux indicateurs de texture extraits des images satellites. Sa mise en oeuvre sur des données associées au Mali pose les premières bases d’une méthode originale d'extraction de motifs séquentiels à partir de données complexes.

• Opportunistic Adaptation Knowledge Discovery. F. Badra, A. Cordier and J. Lieber (2009)
  8^th International Conference on Case-Based Reasoning, ICCBR'09.
  abstract | paper | slides | bibtex

  @inproceedings{iccbr-BadraCL09,
  author    = {Fadi Badra and Am{\'e}lie Cordier and Jean Lieber},
  title     = {Opportunistic Adaptation Knowledge Discovery},
  booktitle = {Case-Based Reasoning Research and Development, 8th International Conference on Case-Based Reasoning, 
  	     ICCBR 2009, Seattle, WA, USA, July 20-23, 2009, Proceedings},
  year      = {2009},
  pages     = {60-74}
  }
  			

  Adaptation has long been considered as the Achilles' heel of case-based reasoning since it requires some domain-specific knowledge that is difficult to acquire. In this paper, two strategies are combined in order to reduce the knowledge engineering cost induced by the adaptation knowledge (CA) acquisition task: CA is learned from the case base by the means of knowledge discovery techniques, and the CA acquisition sessions are opportunistically triggered, i.e., at problem-solving time.

• Knowledge Acquisition and Discovery for the Textual Case-Based Cooking system WikiTaaable. F. Badra, J. Cojan, A. Cordier, J. Lieber, T. Meilender, A. Mille, P. Molli, E. Nauer, A. Napoli, H. Skaf-Molli, Y. Toussaint (2009)
  Computer Cooking Contest Workshop at the 8^th International Conference on Case-Based Reasoning, ICCBR'09.
  abstract | paper | bibtex

  @inproceedings{iccbr-wikitaaable,
  title={Knowledge Acquisition and Discovery for the Textual Case-Based Cooking system \textsc{WikiTaaable}},
  author={F. Badra and J. Cojan and A. Cordier and J. Lieber and T. Meilender and A. Mille and P. Molli and E. Nauer
          and A. Napoli and H. Skaf-Molli and Y. Toussaint},
  booktitle={8th {I}nternational {C}onference on {C}ase-{B}ased {R}easoning - {ICCBR} 2009, Workshop Proceedings },
  editor = {Lorraine McGinty and David C. Wilson},
  year=2009
  }
  			

  The textual case-based cooking system WikiTaaable participates to the second Computer cooking contest (CCC). It is an extension of the Taaable system that has participated to the first CCC. Wikitaaable's architecture is composed of a semantic wiki used for the collaborative acquisition of knowledge (recipe, ontology, adaptation knowledge) and of a case-based inference engine using this knowledge for retrieving and adapting recipes. This architecture allows various modes of knowledge acquisition for case-based reasoning that are studied within the Taaable project. In particular, opportunistic adaptation knowledge discovery is an approach for interactive and semi-automatic learning of adaptation knowledge triggered by a feedback from the user.

• Modeling adaptation of breast cancer treatment decision protocols in the Kasimir project. J. Lieber, M. d'Aquin, F. Badra and A. Napoli (2008)
  in Applied Intelligence, 28(3):261--274.
  abstract | paper | bibtex

  @article{apin-LieberdBN08,
    author    = {J. Lieber and M. d'Aquin and F. Badra and A. Napoli},
    title     = {Modeling adaptation of breast cancer treatment decision
                 protocols in the Kasimir project},
    journal   = {Applied Intelligence},
    volume    = {28},
    number    = {3},
    year      = {2008},
    pages     = {261-274}
  }
  			

  Medical decision protocols constitute theories for health-care decision making that are applicable for “ standard ” medical cases but have to be adapted for the other cases. This holds in particular for the breast cancer treatment protocol studied in the Kasimir research project. Protocol adaptations can be seen as knowledge-intensive case-based decision support processes. Some examples of adaptations that have been performed by oncologists are presented in this paper. Several issues are then identified that need to be addressed while trying to model such processes, namely: the complexity of adaptations, the lack of relevant information about the patient, the necessity to take into account the applicability and the consequences of a decision, the closeness to decision thresholds, and the necessity to consider some patients according to different viewpoints. As handling these issues requires some additional knowledge, which has to be acquired, different methods are presented that perform adaptation knowledge acquisition either from experts, or in a semi-automatic manner. A discussion and a conclusion end the paper.

• EdHibou: A Customizable Interface for Decision Support in a Semantic Portal F. Badra, M. d'Aquin, J. Lieber and T. Meilender (2008)
  Poster/Demo at the 7^th International Semantic Web Conference, ISWC'08
  abstract | paper | poster | bibtex

  @inproceedings{semweb-BadradLM08,
    author    = {F. Badra and M. d'Aquin and J. Lieber and T. Meilender},
    title     = {EdHibou: a Customizable Interface for Decision Support in
                 a Semantic Portal},
    booktitle = {Proceedings of the Poster and Demonstration Session at the
                 7th International Semantic Web Conference (ISWC2008), Karlsruhe,
                 Germany, October 28},
    year      = {2008}
  }
  			

  The Semantic Web is becoming more and more a reality, as the required technologies have reached an appropriate level of maturity. However, at this stage, it is important to provide tools facilitating the use and deployment of these technologies by end-users. In this paper, we describe EdHibou, an automatically generated, ontology-based graphical user interface that integrates in a semantic portal. The particularity of EdHibou is that it makes use of OWL reasoning capabilities to provide intelligent features, such as decision support, upon the underlying ontology. We present an application of EdHibou to medical decision support based on a formalization of clinical guidelines in OWL and show how it can be customized thanks to an ontology of graphical components.

• Taaable: Text Mining, Ontology Engineering, and Hierarchical Classification for Textual Case-Based Cooking. F. Badra, R. Bendaoud, R. Bentebibel, P.-A. Champin, J. Cojan, A. Cordier, S. Desprès, S. Jean-Daubias, J. Lieber, T. Meilender, A. Mille, E. Nauer, A. Napoli and Y. Toussaint (2008)
  Computer Cooking Contest Workshop at the 9^th European Conference on Case-Based Reasoning, ECCBR'08.
  abstract | paper | bibtex

  @inproceedings{ewcbr-taaable,
    author    = {Fadi Badra and Rokia Bendaoud and Rim Bentebibel and Pierre-Antoine Champin and Julien Cojan and
                 Am{\'e}lie Cordier and Sylvie Despr{\`e}s and  St{\'e}phanie Jean-Daubias and Jean Lieber and
                 Thomas Meilender and Alain Mille and Emmanuel Nauer and Amedeo Napoli and Yannick Toussaint},
    title     = {\textsc{Taaable}: Text Mining, Ontology Engineering, and Hierarchical Classification for Textual Case-Based Cooking},
    booktitle = {ECCBR 2008, The 9th European Conference on Case-Based Reasoning,
                 Trier, Germany, September 1-4, 2008, Workshop Proceedings},
    year      = {2008},
    pages     = {219-228},
    editor    = {Martin Schaaf}
  }
  			

  This paper presents how the Taaable project addresses the textual case-based reasoning challenge of the CCC, thanks to a combination of principles, methods, and technologies of various fields of knowledge-based system technologies, namely CBR, ontology engineering (manual and semi-automatic), data and text-mining using textual resources of the Web, text annotation (used as an indexing technique), knowledge representation, and hierarchical classification. Indeed, to be able to reason on textual cases, indexing them by a formal representation language using a formal vocabulary has proven to be useful.

• Representing Case Variations for Learning General and Specific Adaptation Rules. F. Badra and J. Lieber (2008)
  STAIRS 2008- Proceedings of the Fourth Starting AI Researchers' Symposium, Patras, Greece, 21-25 July, 2008
  abstract | paper | slides | bibtex

  @inproceedings{stairs-BadraL08,
    author    = {Fadi Badra and Jean Lieber},
    title     = {Representing Case Variations for Learning General and Specific Adaptation Rules},
    booktitle = {STAIRS 2008   - Proceedings of the Fourth Starting AI Researchers'
                 Symposium, Patras, Greece, 21-25 July, 2008},
    year      = {2008},
    publisher = {IOS Press},
    editor    = {Amedeo Cesta and Nikos Fakotakis},
    series    = {Frontiers in Artificial Intelligence and Applications},
    volume    = {179},
    pages     = {1-11}
  }
  			

  Adaptation is a task of case-based reasoning systems that is largely domain-dependant. This motivates the study of adaptation knowledge acquisition (AKA) that can be carried out thanks to learning processes on the variations between cases of the case base. This paper studies the representation of these variations and the impact of this representation on the AKA process, through experiments in an oncology domain.

• Case Base Mining for Adaptation Knowledge Acquisition. M. d'Aquin, F. Badra, S. Lafrogne, J. Lieber, A. Napoli and L. Szathmary (2007)
  20^th International Joint Conference on Artificial Intelligence, IJCAI'07
  abstract | paper | bibtex

  @inproceedings{IJCAI07,
    author = {M. d'Aquin and F. Badra and S. Lafrogne and J. Lieber and A. Napoli and L. Szathmary},
    title = {Case Base Mining for Adaptation Knowledge Acquisition},
    booktitle = {Proceedings of the International Conference on Artificial Intelligence, IJCAI'07},
    year = {2007},
    pages = {750--756}
  }
  			

  In case-based reasoning, the adaptation of a source case in order to solve the target problem is at the same time crucial and difficult to implement. The reason for this difficulty is that, in general, adaptation strongly depends on domain-dependent knowledge. This fact motivates research on adaptation knowledge acquisition (AKA). This paper presents an approach to AKA based on the principles and techniques of knowledge discovery from databases and data-mining. It is implemented in CabamakA, a system that explores the variations within the case base to elicit adaptation knowledge. This system has been successfully tested in an application of case-based reasoning to decision support in the domain of breast cancer treatment.

• Adaptation Knowledge Discovery from a Case Base. M. d'Aquin, F. Badra, S. Lafrogne, J. Lieber, A. Napoli and L. Szathmary (2006)
  17^th European Conference on Artificial Intelligence, ECAI'06
  abstract | paper | poster | bibtex

  @article{aquin-ecai06,
    author    = {Mathieu d'Aquin and Fadi Badra and Sandrine Lafrogne and Jean Lieber and Amedeo Napoli and Laszlo Szathmary},
    title     = {Adaptation Knowledge Discovery from a Case Base},
    journal   = {ECAI 2006, 17th European Conference on Artificial Intelligence,
                 August 29 - September 1, 2006, Riva del Garda, Italy, Including
                 Prestigious Applications of Intelligent Systems (PAIS 2006),
                 Proceedings},
    publisher = {IOS Press},
    editor    = {Gerhard Brewka and Silvia Coradeschi and Anna Perini and Paolo Traverso},
    series    = {Frontiers in Artificial Intelligence and Applications},
    volume    = {141},
    year      = {2006}
  }
  			

  In case-based reasoning, the adaptation step depends in general on domain-dependent knowledge, which motivates studies on adaptation knowledge acquisition (AKA). CABAMAKA is an AKA system based on principles of knowledge discovery from databases. This system explores the variations within the case base to elicit adaptation knowledge. It has been successfully tested in an application of case-based decision support to breast cancer treatment.

National (french) journals, conferences and workshops

• Découverte opportuniste de connaissances d'adaptation. F. Badra, A. Cordier and J. Lieber (2009)
  17^ème atelier sur le raisonnement à partir de cas - RàPC'09.
  abstract | paper | slides | bibtex

  @inproceedings{rapc09-badra, 
  title={D\'ecouverte opportuniste de connaissances d'adaptation},  
  author={F. Badra and A. Cordier and J. Lieber}, 
  booktitle={17{\`e}me atelier sur le raisonnement {\`a} partir de cas - {R}{\`a}{PC}-09 }, 
  pages={23--34}, 
  address={{P}aris {F}rance },
  editor={B\'eatrice Fuchs and Amedeo Napoli},
  year={2009},
  }			
  			

  L'étape d'adaptation est souvent considérée comme le talon d'Achille du raisonnement à partir de cas car elle nécessite des connaissances spécifiques au domaine d'application qui sont difficiles à acquérir. Dans ce papier, deux stratégies sont combinées pour faciliter la tâche d'acquisition de connaissances d'adaptation : les connaissances d'adaptation sont apprises à partir de la base de cas par des techniques d'extraction de connaissances, et l'acquisition de connaissances d'adaptation est déclenchée de manière opportuniste au cours d'une session particulière de résolution de problèmes.

• Familles génératrices de règles d'adaptation pour assister leur acquisition semi-automatique. M. Tixier, F. Badra and J. Lieber (2008)
  19^èmesJournées Francophones d'Ingénierie des Connaissances (IC'08).
  abstract | paper | bibtex

  @inproceedings{f-ic-TixierBL08,
    author    = {M. Tixier and F. Badra and J. Lieber},
    title     = {Familles g{\'e}n{\'e}ratrices de r{\`e}gles d'adaptation
                 pour assister leur acquisition semi-automatique},
    booktitle = {Actes d'IC  2008: Ing{\'e}nierie des connaissances 2008 (Proceedings  
                 of the 19th French Knowledge Engineering Conference), Nancy,
                 France, June 18-20, 2008},
    year      = {2008},
    pages     = {225-236}
  }		
  			

  L'acquisition de connaissances d'adaptation, notamment dans sa dimension automatique, ouvre de grandes perspectives pour l'étape critique en raisonnement à partir de cas que constitue l'adaptation. Le système CabamakA s'appuie sur des techniques symboliques de fouille de données pour extraire des règles d'adaptation candidates. Cet article présente une approche pour réduire le nombre de telles règles à présenter à l'analyste pour validation. L'idée est de s'appuyer sur une opération de composition de ces règles et de chercher une famille génératrice pour cette composition.

• Une approche pour représenter les variations entre cas - Vers une application à l'extraction de connaissances d'adaptation. F. Badra and J. Lieber (2007)
  15^ème atelier sur le raisonnement à partir de cas - RàPC'07.
  abstract | paper | bibtex

  @inproceedings{rapc07-badra, 
  title={{U}ne approche pour repr{\'e}senter les variations entre cas
                    --- {V}ers une application {\`a} l'extraction de
                    connaissances d'adaptation},  
  author={F. {B}adra and J. {L}ieber},
  booktitle={15{\`e}me atelier sur le raisonnement {\`a} partir de cas - {R}{\`a}{PC}-07 }, 
  pages={47--56 }, 
  address={{G}renoble {F}rance },
  editor={{A}m{\'e}lie {C}ordier },
  year={2007},
  }		
  			

  Cet article propose une méthode pour représenter les variations entre deux cas dans le cadre d'une représentation des cas par des couples attributs-ensembles de valeurs (ce qui généralise le formalisme, classique dans le cadre du raisonnement à partir de cas, des couples attributs-valeurs). Pour effectuer une telle représentation, on met en évidence tout d'abord les attributs qui ne sont pas communs aux deux cas puis, pour les autres, on compare les ensembles de valeurs par des ensembles de relations binaires. De telles relations sont spécifiées par le concepteur du système de raisonnement à partir de cas sur la base de ce qui est pertinent pour « passer » d'un cas à un autre. Des exemples de telles relations sont proposés. Ce travail doit être appliqué à l'extraction de connaissances d'adaptation par fouille de la base de cas : une telle fouille s'appuie sur la représentation des variations entre cas de la base. L'expressivité des règles d'adaptation ainsi extraites sera fonction de l'expressivité de la représentation des variations entre cas.

• Extraction de connaissances d'adaptation par analyse de la base de cas. F. Badra and J. Lieber (2007)
  7^èmes Journées Extraction et Gestion des Connaissances (EGC'07).
  abstract | paper | slides | bibtex

  @inproceedings{EGC07,
  title =	{Extraction de connaissances d'adaptation par l'analyse de la base de cas},
  author =	"F. Badra and J. Lieber",
  Booktitle =	"Extraction et gestion des connaissances ({EGC}'2007),
  		 Actes des septi{\`e}mes journ{\'e}es Extraction et
  		 Gestion des Connaissances, Namur, Belgique, 23-26 janvier
  		 2007, 2 Volumes",
  publisher =	"C{\'e}padu{\`e}s-{\'E}ditions",
  year = 	"2007",
  pages =	"751--760",
  series =	"Revue des Nouvelles Technologies de l'Information"
  }
  		
  			

  En raisonnement à partir de cas, l'adaptation d'un cas source pour résoudre un problème cible est une étape à la fois cruciale et difficile à réaliser. Une des raisons de cette difficulté tient au fait que les connaissances d'adaptation sont généralement dépendantes du domaine d'application. C'est ce qui motive la recherche sur l'acquisition de connaissances d'adaptation (ACA). Cet article propose une approche originale de l'ACA fondée sur des techniques d'extraction de connaissances dans des bases de données (ECBD). Nous présentons CabamakA, une application qui réalise l'ACA par analyse de la base de cas, en utilisant comme technique d'apprentissage l'extraction de motifs fermés fréquents. L'ensemble du processus d'extraction des connaissances est détaillé, puis nous examinons comment organiser les résultats obtenus de façon à faciliter la validation des connaissances extraites par l'analyste.

Other publications

• Utilisation de techniques de fouille de données pour enrichir une ontologie d'objets géographiques. F. Badra (2005)
  Rapport de Master Recherche informatique spécialité Extraction de connaissances à partir de Données (Equipe de Recherche en Ingénierie des Connaissances, Lyon).
  abstract | manuscrit | site Web du laboratoire ERIC

  La recherche sur les ontologies suscite un intérêt croissant dans le domaine du traitement de l'information géographique. Celles-ci apparaissent comme un composant essentiel pour réaliser l'intégration de sources de données hétérogènes et faire face à la considérable augmentation de la quantité de données d'observation de la Terre aujourd'hui disponibles. Cette étude s'inscrit dans le projet multidisciplinaire FoDoMuST, qui vise à extraire et qualifier des objets urbains à partir d'images satellites multi-spectrales et multi-résolution. Un objet urbain est un concept de la géographie urbaine, discipline scientifique qui étudie les dimensions spatiales du phénomène urbain et dont l'objet d'étude est la ville. Nous nous sommes concentrés essentiellement sur les aspects conception et implémentation d'une ontologie d'objets urbains. Notre démarche a abouti à la proposition d'un modèle conceptuel de l'ontologie ainsi qu'à une étude de son opérationnalisation et des perspectives de son enrichissement par des processus d'extraction de connaissances à partir de données.

• Comparative Study on Multi-Class Composite N-grams Applied to English and Japanese. F. Badra (2003)
  Rapport de projet de fin d'études de la formation d'ingénieur en Télécommunications (Advanced Telecommunications Research Institute, Kyoto, Japon).
  abstract | manuscrit | site Web du laboratoire ATR

  This document will present the research work I conducted in ATR for five months from april to august 2003. My work mainly consisted in an application of Multi-Class Composite N-gram language models, proposed recently by ATR researchers H.Yamamoto, S.Isogai and Y.Sagisaka (Yamamoto, 2001) and only applied to Japanese language so far, to the English language. The purpose of the experiments I conducted was to provide experimental data on such a model for the two languages, and determine to which extend this new technique, which showed good results for Japanese, can be applied as well to the English model. These tests were performed for training corpora of dif- ferent sizes, extracted from the B.T.E.C., and running these experiments on the two languages in the same conditions enabled us to make a comparison between them. The results showed that Multi-Class language models improve conventional Class-Based ones for English too, but with different optimal connectivity information and only for small training corpora.

• Introduction à XML
  Cours donné en 2011 en Master Bioinfo
  cours
• Bases de données relationnelles
  Cours donné en 2011 en Master Bioinfo
  cours
• Introduction aux Web (HTML, CSS)
  Cours donné en 2008 en Licence SMP (Science Mathématique Physique) au CESS d'Épinal
  cours | exo 1 | exo 2
• Introduction aux Web Sémantique
  Intervention en 3^ème année à l''école des Mines de Nancy
  Plan du cours | Partie 1 (Intro) | Partie 2 (Modélisation des données) | Partie 3 (Représentation des connaissances et raisonnement)