Research Statement
I am interested in artificial intelligence, particularly in the background knowledge needed by an AI. I believe the Semantic Web is the key to setting up a large interconnected knowledge base that will allow a new form of intelligence to emerge. I am working along three axes around semantic technologies:
- Ontology alignment: ontology alignment relates to the techniques and processes allowing to construct bridges between heterogeneous ontologies modeling a similar domain. I develop a language called EDOAL: Expressive and Declarative Ontology Alignment Language allowing to represent complex correspondences between the entities of two ontologies. This language allows for a high-level representation of ontology alignments. I have also worked on a SPARQL extension for the transformation of instance data.
- Ontology and correspondence patterns: Anyone who tried to build an ontology will tell it, it is a difficult task. In order to facilitate this task ontology patterns come as reusable pieces of knowledge that can be assembled together in order to build the ontology. I am participating in the effort to construct a comprehensive catalog of ontology patterns: http://www.ontologydesignpatterns.org. I was particularly involved through my PhD thesis in the modeling of correspondence patterns, relating together similar ontology patterns.
- Application to linked-data: The Web of data is an emerging giant interconnected database. It uses semantic technologies and the Web as a support. The Web of data offers an incredible source of data for mining equivalence relations and patterns between ontologies and data. The Web of data also raises new problems related to the scale of the data published. It requires to develop new methods to store, access and manipulate the data. I am applying ontology alignment techniques to the Web of data in order to accelerate its emergence. I particularly concentrate on automating the process of finding equivalent resources. I am also interested in analyzing its structure in order to find regularities.
Projects
Future
- DataLift: DataLift will be a catalyser for data publication on the Web of data.
- KPat: The KPAT project will develop a new science of knowledge patterns.
Current
- NeOn: A European project where we reseachr on the theme of networked ontologies. I am involved in constructing alignments and managing contexts between network ontologies.
Past
- Web content: a French project that did research on a semantic Web platform. I was involved in the ontology matching part of the platform.
- EASAIER: A EU project on Enabling Access to Sound Archives through Integration, Enrichment and Retrieval. I was leader of the workpackage 2: Semantic infrastructure. This was the first occasion I had to experiment the use of semantic technologies in a practical scenario, it gave very interesting results.
- Knowledge Web: this EU network of excellence provided many results on semantic Web technologies. I was a member of the workpackage 2.2: Heterogeneity.
- SEKT Semantically Enabled Knowledge Technologies. This EU project aimed to develop new techniques for knowledge management using semantic technologies. I was in the workpakage 4: ontology mediation.
- DIP Data, Information, and Process Integration with Semantic Web Services, a EU project oriented toward Semantic Web services.
Tools
This section presents a few tools I have been developing over the last years.
- RDF-AI: A data interlinking tool developed in collaboration with Yanbin Liu of Jilin university.
- RDFS2SKOS: A tool for converting from RDFS ontologies to SKOS vocabularies.
- SEKT Mediator: an ontology mediation tool developed in collaboration with the company Ontoprise in the context of the SEKT project. The tool is not available but a video can be seen there.
- EASAIER sound-acces architecture: This system allows to annotate music and sounds by extracting metadata automatically using signal analysis tools. This too is based on a semantic reposirtory. It was developed in the context of the EASAIER project.
- OMWG Mapping APIw: This API allos to manipulate statements of the ontology alignment language I developed during my PhD. It contains a parser based on SableCC, and an object model giving basic manipulation and serialization of the language statements.
- DIP Mapping and Merging Tool: An ontology integration tool developed n the context of the DIP EU project.
- Tempo: Tempo is a chess engine I developed during my master at the university Paul-Sabatier in Toulouse. Tempo has the particularity of playing using a neural network. The structure of the network was found using a genetic algorithm.