Abstract
Pervasive computing envisions a world where "invisible" computers seamlessly assist users in their daily lives. However, achieving this requires overcoming significant hurdles in interoperability and context awareness among heterogeneous Web services. This paper proposes the SeMEther framework, which leverages Semantic Web technologies to automatically compose services into context-aware "service flows." By distinguishing between syntactic, semantic, and pragmatic knowledge, SeMEther enables systems to evaluate the social appropriateness of service execution, moving beyond mere logical connectivity.
Introduction
The primary goal of pervasive computing is to provide seamless assistance through "invisible" technology. Historically, this vision relied on specialized smart devices with limited interoperability. A major remaining challenge is the dynamic integration of diverse, independently developed Web services based on a user's ever-changing environment.
The core issue is heterogeneity: services use different ontologies and lack a shared understanding of "context." To address this, the authors identify three critical requirements:
- Interoperability: Enabling disparate services to function within a single workflow.
- Knowledge Definition: Distinguishing between structural, logical, and situational service requirements.
- Ontology Integration: Implementing automated, efficient mapping between different data models to facilitate real-time communication.
The SeMEther Framework
The authors introduce SeMEther, a framework that utilizes a three-tiered knowledge approach for service description:
- Syntactic Knowledge: Structural rules (e.g., data formats like 10-digit phone numbers).
- Semantic Knowledge: Logic-based rules often requiring external verification (e.g., checking if a phone number is active).
- Pragmatic Knowledge: Contextual rules regarding social appropriateness (e.g., suppressing non-emergency calls late at night).
The architecture includes several key components:
- Service Composition Subsystem (WSCS): This includes a crawler for service discovery, an editor for ontology mapping, and a composer that utilizes Jena rules and OWL-S to build dynamic workflows.
- Context Management: A Knowledge Base (KB) Handler tracks user spatial position and status to trigger appropriate location-based services.
- Unified Information Base (UIB): Acts as a bridge between disparate data sources and standard ontologies.
Ontology Integration and Datasets
A novel component of SeMEther is OnInt (Ontology Integration), a two-level approach using a Terminological Knowledge Base (TKB). It first maps concepts to high-level "semantic types" to prune the search space, then employs bigram string matching and relationship mapping to calculate similarity scores efficiently.
Rather than traditional machine learning datasets, the framework extends established Semantic Web ontologies to provide a foundation for pervasive computing.
As shown in the table above, the system utilizes CONON as its upper ontology, while leveraging SOUPA for time and location, and W3C DISelect for device modeling.
Key Insights: Pragmatics and Efficiency
The most significant conceptual contribution is the focus on Pragmatics. While most service discovery focuses on whether a service can run, SeMEther asks if it should run based on social context. This shift is vital for making pervasive computing truly "invisible" and non-intrusive.
Furthermore, the system achieves efficiency through its hierarchical ontology mapping. By using a two-level TKB, the computational cost of real-time integration is significantly reduced. Finally, by moving context storage and composition to the Web, the framework reduces reliance on expensive, specialized hardware, allowing standard mobile devices to participate in the ecosystem.
Results and Future Work
The SeMEther prototype was evaluated using a "Buddy" service proof-of-concept. The system successfully detected nearby contacts via Bluetooth and dynamically adapted its behavior. For instance, if a user was in a meeting (determined via schedule and location), the system automatically suppressed alerts and forwarded messages to email.
While successful, the current centralized KB approach may face scaling issues. Future work will explore distributed knowledge bases, negotiation rules for "fuzzy" service matching, and more sophisticated user preference modeling to further refine service requests.
Summary updated