Adaptive Private Profiling for Large-scale Open Information Systems

Information Systems, such as the Web, Peer-to-Peer networks, Grid Computing, Information Retrieval, have to cope with an increasing number and the heterogeneity of their components, and with a continual environmental dynamics. Several general problems remain in these complex systems: (i) the adaptation to the environmental dynamics, (ii) the adaptation to the needs, the centres of interest and the preferences of the real services, (iii) a better mapping of a requirement and an offer by taking into account multiple criteria dependent on the current field.
This observation has direct effects on the Quality of Service (QoS) given by such systems. This dynamic, combined with the inherent geographical and functional distribution of information systems, makes usual adaptation techniques unsuitable because they are global or dependent of the intended global function realised by the system. Consequently, new paradigms based on adaptive and local approaches need to be defined in that sense.
We propose a self-organising technology to optimize the global QoS. This approach allows real-time learning from observations and experiences to adapt information systems to their contexts. It also gives more personalised and user-friendly services to an end-user.
This project aims to apply this self-organising technology to several fields, such as Peer-to-Peer, e-tourism or Information Retrieval.

Strategic objectives

• enabling discovery, advertising, personalisation and dynamic composition of services
• guaranteeing an optimal quality of service and privacy for end-users
• insuring behavioural stability in self-organising complex systems

Strategic expected achievements
A cooperative agent shell for end-users and services allowing an emergent and optimal collective behaviour of self-organising complex systems.
This shell will be proofed in several concrete application domains such as e-tourism, peer-to-peer and information retrieval.

Proposal Outline
The objectives of this project are triple to cope with the inherent complexity of Open Information Systems by giving to each component of such systems, the way to locally and autonomously adapt to unexpected events that can occur in their environment. To deal with the inherent complexity of Open Information systems and to optimize their global QoS, we need to:

• Instantiate a self-organising technology to allow the best user/services mapping according to a need while taking into account the environmental dynamics such as breakdown or appearance of services.
• Define a method enabling the construction of an adaptive and contextual profile about each involved entity (service or user). Thus, representations owned by each entity have to be local and they also have to be personalised according to interactions frequency and context.
• Apply the project to really Open Information Systems. For that we need to construct terminological networks emerging from the history of exchanges between involved entities. These terminological networks should be used for profiles construction as well as request reformulation and results merging for end-users.
• Provide to each involved entity a local and adaptive multi-criteria method to sort, according to owned profiles, the relevant services. This part should have an immediate impact on the QoS.
• Develop a toolbox to observe the Information Systems dynamics and the convergence of the QoS towards the optimum.

Several interesting aspects are directly induced by the proposed local adaptive and domain-independent approach:

• A local privacy management. As behaviours of each involved entity are managed in a strictly local way, by the entities themselves and according to their neighbourhoods, a global vision of the system content is useless.
• A graceful degradation. As each entity is locally endowed by adaptive and cooperative behaviours it is able to autonomously take into account unexpected situations (overload, breakdown…). This is also an important criterion for the QoS.

Local and adaptive behaviours of entities should be totally independent of any application domain and should be reused in any Open Information Systems. That is the reason why we want to apply them to several fields such as peer-to-peer, information retrieval or e-tourism which are typical fields for the proposed technology.

Open Information Systems, Quality of Service, adaptive profiling, privacy, self-organising systems

• Know-how on multi-criteria management, on design of interactive user interface including ergonomic aspects.
• Usable ontologies in a given field (possibly e-tourism) or partners interested in applications needing the dynamic construction of terminological networks or ontologies.
• Expertise in peer-to-peer and e-tourism applications, test beds.

• Partners to develop software engineering tools which will constitute the test bed.
• Company partners to conduct large-scale real world experiments and to provide existing services.
• Both industrials and academics have to contribute to the requirements definition.

• Partners that are manufacturer/provider of e-tourism services or other relevant large-scale domains.
• Academics or companies for developing the platform.

Yes