Conférenciers invités

Valentina Beretta

Position actuelle : Chercheuse post-doctorale à l'institut de recherche pour le développement (IRD)

Titre de la présentation : Data Veracity Assessment: How a-priori Knowledge can Improve Truth Discovery Models

Résumé :

Data veracity is one of the main issues regarding Web data. Automatically identifying reliable information is a critical task within a knowledge extraction process, in particular if resulting knowledge bases (KB) are devoted to be used in decision processes. Truth discovery models address it comparing information provided by multiple sources. They assume that true information is provided by reliable sources and, vice versa, reliable sources provide true information. In this way, they are able to identify which are the true claims among a set of conflicting ones. To the best of our knowledge, existing methods do not consider prior knowledge that can be extracted from an ontology. Based on the type of information that is considered, important suggestions that can facilitate the identification of true claims can be recognized. In our study, we analyze the impact of incorporating into truth discovery model the information associated with concepts' hierarchy and frequent patterns that are extracted from KB using association rule learning techniques. Empirical experiments on synthetic and real world datasets show advantages and disadvantages of proposed models.

Allel Hadjali

Position actuelle : Professeur, LIAS/ENSMA, Poitiers

Page Web : https://www.lias-lab.fr/members/allelhadjali

Titre de la présentation : A Panorama of Data Uncertainty Models

Résumé :

Uncertainty is one major aspect of imperfection of data. At least three causes of this kind of imperfection can be distinguished: (i) variability of phenomena which prevents to predicate their behavior, (ii) lack of information, (iii) conflicting/evolving information. Several models to represent uncertainty exist in the literature. It is well known that each model is insufficient to handle all facets of uncertainty. Some models are more general than others, some are more fitted to a given situation than others, some are more mature than others, some have interpretations better fitted to a particular situation or problem, some dispose of more convenient tools than others. In this talk, the aim is not to show "which is the best model" but rather "when, where, why and how each model should be used?" Illustrations borrowing from Databases field are provided.