DRUID considers models and algorithms for the management and qualification of uncertain, participative, interlinked and large-scale data (social networks, sensor networks, Web streams, crowdsourcing, …).

The objective of the DRUID team is to provide models and algorithms for the annotation and management of interlinked data and sources at a large scale. We consider three main goals:

  1. To propose well-founded models for interlinked data and, more importantly, interlinked data sources (for example, orchestrating users and tasks in crowdsourcing platforms),
  2. To develop theories for the qualification of such data and sources in terms of reliability, certainty, provenance, economical value, trust …
  3. To implement systems that are proof-of-concepts of these models and theories. In particular we would like to demonstrate that these systems can overcome specific key problems in real-world applications, such as scalable data qualification and data adaptation to the final users.

More concretely, we would like to address the following challenges:
to develop integrated and scalable analysis tools for participants in social networks, that encompass the semantics of communications between users user influence, and other aspects.

  • to extend existing crowdsourcing platforms with fine user profiles, team building, complex task management, with application for e-science or e-government (smart cities).
  • to develop reliability assessment techniques for large sensor networks (uncertainty), heterogeneous data sources or Linked Open Data (quality), or microblog conversations (misinformation).
  • to adapt data to its use (data visualization, accessibility of information).

Scientific foundations

For data management we will naturally elaborate on classical techniques: finite model theory, complexity theory, approximation algorithms, declarative or algebraic languages, execution plans, costs models, indexing.

For data qualification, our first focus will be on uncertainty.  Many frameworks are available, but all are based on the theories of uncertainty. We have to
differentiate two main aspects of imperfection: uncertainty and imprecision. The theory of belief functions will be firstly considered for its richer representation of uncertainty and imprecision compared to probability theory and its higher ability to combine pieces of information (for the task of information fusion).