Personalized Information Retrieval (PIR) exploits the user's data in order to refine the retrieval, like for instance when users with different backgrounds may express different information needs with the same query. However, this additional source of information is not supported by the classical Information Retrieval (IR) process. In order to overcome this limit, we propose to generate the user profile out from his profile and social data. Then, we introduce several Personalized Information Retrieval models which integrate this profile at the querying step, allowing to personalize the search results. We study several combinations of the initial user's query with his profile. Furthermore, we present a PIR test collection that we built from the social bookmarking network Delicious, in order to evaluate our PIR models. Our experiments showed that the PIR models improve the retrieval results.

Social Information Retrieval (SIR) exploits the user's social data in order to refine the retrieval, for instance in the case where users with different backgrounds may express different information needs as a same textual query. However, this additional source of information is not supported by the classical IR process. In this article, we propose an approach to generate the user profile from his social data. This generated profile is integrated within a SIR model allowing to personalize the list of documents returned to the user.

If the clustering task is widely studied both in graph clustering and in non supervised learning, combined clustering which exploits simultaneously the relationships between the vertices and attributes describing them, is quite new. In this paper, we present different scenarios for this task and, we evaluate their performances and their results on a dataset, with ground truth, built from several sources and containing a scientific social network in which textual data is associated to each vertex and the classes are known. We argue that, depending on the kind of data we have and the type of results we want, the choice of the clustering method is important and we present some concrete examples for underlining this.

In this paper, we present different combined clustering methods and we evaluate their performances and their results on a dataset with ground truth. This dataset, built from several sources, contains a scientific social network in which textual data is associated to each vertex and the classes are known. Indeed, while the clustering task is widely studied both in graph clustering and in non supervised learning, combined clustering which exploits simultaneously the relationships between the vertices and attributes describing them, is quite new. We argue that, depending on the kind of data we have and the type of results we want, the choice of the clustering method is important and we present some concrete examples for underlining this.

In this paper, we present a hyperdocument model taking into account the essential aspects of information on the Web: content, composition (logical structure) and non-linear reading (hypertext structure). We have developed a Structured Information Retrieval System (SIRS) based on this model. Its phases of indexing and querying are based on a “reading paths” point of view of the Web: a Web site is considered as a set of potential reading paths, instead of a set of atomic and flat pages. We have developed an specific algorithm to index the reading paths. We present some experiments aiming at evaluating the interest of our indexing process of reading paths.

In databases or in the World Wide Web, many documents are in a structured format (e.g. XML). We propose in this article to extend the classical IR probabilistic model in order to take into account the structure through the weighting of tags. Our approach includes a learning step in which the weight of each tag is computed. This weight estimates the probability that the tag distinguishes the terms which are the most relevant. Our model has been evaluated on a large collection during INEX IR evaluation campaigns.

The growth of the Web has posed new challenges for Information Retrieval (IR). Most of the current systems are based on traditional models, which have been developed for atomic and independents documents and are not adapted to the Web. A promising research orientation consists of studying the impact of the Web structure on indexing. The HyperDocument model presented in this article is based on essential aspects of information comprehension: content, composition and linear/non-linear reading.