GUILLAUME ERETEO

"I hope that while so many people are out smelling the flowers, someone is taking the time to plant some." Herbert Rappaport

PhD thesis Subject:

Social network analysis is a key technique in sciences aimed to analyse structures, roles and position of resources linked by properties in a network. In most sciences, like sociology, the studied resources are persons linked by one or more human relationship. These relationships can have the same type (e.g. 'knows'), but in general relations are of different types. Typically, in an acquaintance network we have family links, friendship, business contact, etc. Resources can also be of different types like woman, teacher, and even organisation or course, etc. Social network analysis can be applied to understand social interactions with physical or conceptual entities. Web 2.0 is a good example where persons are explicitly linked through social networking sites but also implicitly with collaborative platform like wikis, blogs and social bookmarking sites. They share interests, opinions and experiences. This network is composed of different kinds of documents (web pages, post files, multimedia, etc) manipulated by persons. There are links like 'knows' ( and all its possible inheritance) between two persons, but also links like 'author' between a document and a person, 'has tag' between a bookmark and a tag, etc. Thus the web is a huge graph including huge social networks, with different type of resources linked with different types of properties. The objective of these PHD is to applied semantic web technologies to social network analysis.

Represent, annotate and make understandable uses and profiles with semantic web technologies

The first step of this work is to represent social networks with semantic web standards. Such social network is developed by people using web applications, almost as a side effect. It captures persons' profiles, their different online accounts and their multiple usages on the web. We will have to adapt existing standards (foaf, sioc, scot, hCard, etc.), extend them if necessary and manage their intersection in order to enable their integration. We will pay special attention to providing seamless and non intrusive means to capture these annotations and make them understandable and usable by applications. Capture methods as well exchange and aggregation formalisms will be proposed. Finally, the crucial and difficult problem of confidentiality will have to be taken into account. Private and public annotations need to be distinguished and it will be necessary to provide profile access control, using semantic exploitation of these annotations.

Bring social network analysis methods to semantic web technologies in order to detect social structure, roles and positions among users of web resources

By their nature, semantic web annotations, expressed in RDF with ontologies support, form an oriented types (labelled) graph. Nodes and vertices of this kind of graph can be of different types. Ontologies and inferring languages enable us to deduce new annotations from existing ones which in turn add others nodes and vertices to the graph. The main method to analyse social network is graph theory. Lots of algorithms are available to detect social structures, roles and positions. One of the goals of this PHD is to apply these algorithms to social networks that are described with semantic web technologies. Another objective is to extend existing algorithms or create new ones in order to detect interest communities and manage their life cycle taking into account not only the graph structure but also the semantics of the ontological primitives used to label it. Semantic web also provide a querying language, SPARQL, designed to query semantic web annotations. We will semantically represent graphs and investigate the way to abstract graph algorithms with high level annotation querying.