Dr. Salma [email protected]/in/salmanajar/
Salma NajarManuele Kirsch-PinheiroCarine Souveyet
Pervasive Information System (PIS)•Integration of IS in dynamic and heterogeneous environment•Context-awareness and user’s needs satisfaction•Predictable and expected behavior
Pervasive Environment•Integration of new invisible technologies in the daily life
Information System•User’s needs satisfaction•Controllable and predictable
Transparency? Proactivity?Context-Awareness?
Most appropriate
services?
User’s intentions satisfaction?
Innovative approach : User-centred contextual vision of PIS
Intentional approachUser’s intention & intention that
service can satisfy
Contextual approachUser’s current context & service
required context execution ServiceDiscovery
Most appropriate services
Exploitation of the dynamic between intention, context and service
A CONTEXT-AWARE INTENTIONAL SERVICE PREDICTION MECHANISM IN PIS
Transparency
Proactivity
Reduce user’s effort understanding
Hide complexity
User centred Vision
Intention Prediction
Better understanding of user’s future needs and intentions
Answer to user’s needs with a non intrusive way
Context prediction Context-Aware service Recommendation
Approaches - [Sigg et al., 2010]- SCP [Meiners et al., 2010]
- [Abbar et al., 2009]- [Xiao et al., 2010]
Topics - Anticipate user’s next context - Fulfill missing context information
- Proactively propose services to the user
(+) - Provide a proactive behavior - Take into account contextual aspects
(-) - Ignore user’s intentions that emerge in a given context or that hide behind service request - Propose user a service realization, ignoring why it is necessary- Based only on operational variability: the shift between the operational and the intentional layers is not taken into account
• Problem: Non exploitation of the close relation between intention and context in existing prediction and recommendation approaches
Backg
round
Backg
round
Hypothesis: A service prediction mechanism, capable of anticipating user’s intentions in a given context, may improve the overall transparency of PIS.
Research
Problem
Research
Problem
Key
Contributio
n
Key
Contributio
nResults
Experimentation
Results
Experimentation
User situations<Intention, Contexte, Service>
History
Time/
Date
Intention Context Service
T 1 I U1 Cx 1 Sv 1
… .. .. ..T i I U i Cx i Sv i
T n I U n Cx n Sv nTrace ManagementTrace Management
Predicted Intention
ontologies
Prediction ProcessPrediction Process Learning ProcessLearning Process
Prediction
Context-Aware Intentional Semantic Matching Algorithm
Context-Aware Intentional Semantic Matching Algorithm
Markov Chain Algorithm
Markov Chain Algorithm
Context-Aware Intentional Services Prediction Algorithm
Context-Aware Intentional Services Prediction Algorithm
Context-Aware Intentional Services Prediction Mechanism
Prediction Algorithm Quality ResultsPrediction Algorithm Performance
• Evaluation of the Prediction Algorithm• Desktop profile: Machine Intel Core i5 1.3 GHz with 4 GB memory• Dataset
• Extended OWLS-TC2 with intentional and contextual information• Traces database
• Observations• Scalability: Average execution time (performance)• Result Quality: precision and recall
• Polynomial trend of degree three• The number of states increased about
25x, while the execution time has only increased about 2.5x
• More interesting results with a higher quality • Good results depends on:
• Completeness of the ontologies• Setting of the matching threshold
The prediction mechanism allows selecting the most appropriate future service according to the predicted intention in a given context
Intentional approach: more transparent to user
Contextual approach: limits states to those that are valid & executable
[Abbar et al., 2009] Abbar, S., Bouzeghoub, M., and Lopez, S. (2009). Context-Aware Recommender Systems: A Service-Oriented Approach. In 3rd Int Workshop on Personalized Access, Profile Management, and Context Awareness in Databases (PersDB), Lyon, France.[Meiners et al., 2010] Meiners, M., Zaplata, S., and Lamersdorf, W. (2010). Structured Context Prediction: A Generic Approach. In Distributed Applications and Interoperable Systems, F. Eliassen, and R. Kapitza, eds. (Springer Berlin Heidelberg), pp. 84–97.[Sigg et al., 2010] Sigg, S., Haseloff, S., and David, K. (2010). An Alignment Approach for Context Prediction Tasks in UbiComp Environments. IEEE Pervasive Computing, 9(4), pp. 90–97.[Xiao et al., 2010] Xiao, H., Zou, Y., Ng, J., and Nigul, L. (2010). An Approach for Context-Aware Service Discovery and Recommendation. In 2010 IEEE International Conference on Web Services (ICWS), pp. 163–170.
clustering classification
Most appropriate service