Research‎ > ‎

Managing Dynamic Context to Optimize Smart Interactions and Services (IBM CAS Project)

Managing Dynamic Context to Optimize Smart Interactions and Services (3 year project, 2010-2012)

The Personal Web is the people-centric instantiation of the Smart Internet where information systems, services and web content are articulated by users according to their matters of concern. To support the vision of the Personal Web, we have created SmarterContext, a dynamic context management infrastructure that supports end-users in the integration of personal context information using simple but meaningful web interactions according to the specific application domain (e.g., by adding products into shopping wish lists). This context information constitutes the user’s personal context sphere and provides rich knowledge about user’s preferences and situations. The core of our approach is the management of monitoring concerns by implementing feedback loops, where the user acts as the planner of the controller. That is, user-driven web interactions are exploited to adjust users’ context models at runtime. Changes in these context models trigger changes in the monitoring infrastructure to optimize context-awareness and therefore the user’s web experience. The results of this project were validated using two main case studies. The first one focuses on the optimization of the user’s shopping web experience, and the second one on the optimization of adaptation mechanisms to regulate service level agreements (SLA) in large scale service-oriented systems.

This project received IBM CAS Canada Project of the Year 2011 Award and CASCON 2012 Best Paper Award.


  • Hausi A. Müller, Principal Investigator
  • Norha M. Villegas, CAS Student (PhD)
  • Alex Lau, CAS RSM (2010, 2011)
  • Diana Lau, CAS RSM (2012)
  • Joanna Ng, CAS Director
  • Other MSc students who contributed to this project: Priyanka Gupta (University of Victoria); Ishita Jain (University of Victoria);  Sahar Ebrahimi, (University of Victoria); and  Juan C. Munoz (Icesi University, Colombia) 


Objectives for 2010:
  1. To characterize context information along different application domains. 
  2. To investigate context modeling and context managing techniques used in different application domains.
Objectives for 2011:
  1. To define context models for supporting the management of dynamic context as a first-level entity in the Personal Web's data model. 
  2. To design and implement a first prototype of a context management framework to support user-driven web integration in the Smart Internet. 
  3. To integrate the context management framework to the e-commerce case study defined with IBM CAS Canada through the implementation of a web-based prototype.
Objectives for 2012:
  1. To investigate algorithmic techniques for optimizing context reasoning based on contextual RDF graphs.
  2. To investigate suitable methods to guarantee privacy and security of context information in the SmarterContext infrastructure.

Main Components

This project consisted of the following major parts:
  1. Characterization and modeling of context information for smart interactions and smart services. 
  2. Characterization of existing methods and techniques to control and govern the adaptation of smart interactions and smart services due to context changes.
  3. Design of a context modeling mechanism applicable to any domain of the Smart Internet. 
  4. Design and implementation of a context management infrastructure and reasoning engine using WebSphere technologies.
  5. Optimization of the context reasoning engine of SmarterContext using a structural-based approach based on contextual RDF graphs. 
  6. Design and implementation of privacy and security mechanisms required to protect personal context data stored in personal context spheres. 
  7. Validation of the SmarterContext infrastructure for two case studies: (a) smarter commerce and (b) SOA governance. 
General Results
  1. A taxonomy of context types independent of the application domain. 
  2. A set of feature models that characterize required features for context modeling and context management. 
  3. A survey on context modeling and context management approaches.
  4. The SmarterContext ontology: An extensible RDF-based ontology for context representation in any application domain of the Smart Internet. 
  5. The SmarterContext reasoning engine: An extensible set of semantic web rules and their implementation for context reasoning in Smarter Commerce application domains. 
  6. The SmarterContext infrastructure: A context management infrastructure implemented using SCA WebSphere and Rational Application Development technologies. 
  7. The Personal Context Sphere model: A persistent repository that stores personal context information and enables users to integrate and manage their personal context information. Users can integrate third parties to their personal context spheres and authorize them to access their information by defining personalized privacy and security policies. Third parties exploit this information to improve user quality of experience.
  8. SURPRISE, a module of the SmarterContext solution that empowers users with privacy and data security control for the access to their information, stored in Personal Context Sphere repositories. 
Case Studies

We completed the validation of our dynamic context management solution by conducting two case studies. In the first case study, which focused on the personal web, we implemented and evaluated a context-aware deal recommendation system based on the SmarterContext engine. In the second case study, which focused on dynamic SOA governance, we implemented and evaluated a self-adaptation framework that exploits dynamic context management, powered by SmarterContext, to improve context awareness in the governance of service level agreements (SLA) in service-oriented systems.


1.  The SmarterDeals recommendation engine

SmarterDeals implements an algorithm based on collaborative filtering (CF) that analyzes similarities between users and obtain a set of potential relevant deal categories based on context information (i.e., product and service preferences) of similar users. Our recommendation engine improves the accuracy of these potential relevant deal categories by correlating them with context information about the user’s product and service preferences gathered by SmarterContext. SmarterDeals uses location context to filter the recommended categories before their delivery to the user. Our recommendation algorithm contributes to the improvement of the accuracy of recommendations using contextual information provided by our SmarterContext solution. [Read More]

2. The SURPRISE (Smartercontext UseR PRIvacy and SEcurity) module 

SURPRISE is a module of SmarterContext that empowers users with privacy and data security control for the access to their information, stored in Personal Context Sphere repositories. These repositories are defined and maintained by SmarterContext. SURPRISE (i) allows users to configure access permissions to their sensitive personal information to third parties, selectively and with different levels of granularity; (ii) supports changes in these configurations at runtime to add or remove third parties or permissions, and (iii) realizes partial encryption to share non-sensitive data with not explicitly authorized third parties, while protecting user identity. [Read More]

3. A context-aware self-adaptation mechanism supported by SmarterContext

We implemented a prototype of a dynamic SOA governance mechanism that exploits self-adaptation and dynamic context management supported by SmarterContext. This prototype allowed us to demonstrate the effectiveness of our SmarterContext infrastructure to maintain the relevance of its monitoring strategies with respect to changing adaptation goals, thus optimizing the adaptive capabilities of the service-oriented system. [Read More]

4. The SmarterContext engine 

We designed our context reasoning engine based on the features for context handling that we identified in our context modeling and context management conducted in 2010. Our SmarterContext engine supports the inference of implicit contextual facts from explicit contextual data compliant with the SmarterContext ontology. That is, RDF graphs whose nodes and arcs correspond to either types derived from the SmarterContext  ontology or values defined in vocabularies compatible with it. We call these RDF graphs contextual RDF graphs and they constitute personal context spheres. [Read More]

5. The SmarterContext infrastructure 

The SmarterContext infrastructure is a service-oriented self-adaptive system that we developed to manage the lifecycle of dynamic context information. We implemented SmarterContext to: (i) support changes in context management strategies through the self-reconfiguration of both the architecture of its monitoring infrastructure and the business logic of its monitoring conditions. These changes may imply the deployment of new context gatherers and context processing components, or the modification of existing monitoring conditions; (ii) support adaptive reasoning through the addition of new context types, semantic web rules, and pattern-based rules at runtime, without requiring the manual deployment of new software components; and (iii) assist users in gathering, exploiting, maintaining and controlling the access to their personal context information. [Read More]