Mobilizing the Semantic Web with DAML-enabled Web Services

Sheila A. McIlraith, Tran Cao Son, Honglei ZengSemWeb 2001 · OpenReview

Introduction

In the early 2000s, the World Wide Web was primarily a repository for text and images designed for human consumption. Performing complex tasks, such as booking a multi-leg business trip, required users to manually navigate multiple user interfaces. For software agents to automate these tasks, developers typically relied on brittle, site-specific "wrappers" that failed whenever a website's layout changed. This paper proposes a framework to transform the Web into a platform of machine-executable services by combining DAML-enabled ontologies with logic-based programming.

The authors identify three core challenges to achieving "Semantic Web Services":

  1. Automatic Discovery: Enabling agents to find services that meet specific needs.
  2. Automatic Execution: Providing agents with the knowledge of required inputs and expected outputs.
  3. Automatic Composition: Allowing agents to combine independent services (e.g., flight booking, hotel reservation, and mapping) to satisfy complex user requests.

The Semantic Web Framework

The proposed architecture bridges the gap between static web descriptions and active agent reasoning. At its core is an "Agent Broker" that mediates between user-specific agents and various web service ontologies.

Architecture of the Semantic Web Service framework, showing the interaction between agents, brokers, and ontologies.

The framework relies on two primary pillars:

  • DAML+OIL Ontologies: The DARPA Agent Markup Language (DAML) is used to create machine-understandable descriptions of Web services. These descriptions define a service’s IOPEs: Inputs, Outputs, Preconditions (logical states required before execution), and Effects (changes to the world state after execution).
  • ConGolog & Situation Calculus: To provide agents with reasoning capabilities, the authors employ ConGolog, a high-level logic programming language. Web services are modeled as formal actions within the situation calculus.
The Semantic Web language stack, illustrating the hierarchy from XML to DAML-L.

Reasoning and Service Composition

A key innovation is the treatment of Web services as mathematical actions. This allows the application of decades of research in AI planning to the Web. Instead of hard-coding sequences of steps, the authors propose "generic procedures" — high-level task templates (e.g., "Make travel arrangements"). The agent uses deductive machinery to instantiate these templates by searching for available Web services that satisfy both the logical requirements of the task and the user's specific constraints.

User preferences, such as airline choices or travel duration limits, are treated as logical constraints. These are not merely filters applied to the final results; they are used to prune the search tree of possible actions, significantly improving the computational efficiency of service composition.

Results and Implementation

The authors demonstrated the framework through a prototype agent capable of interpreting DAML+OIL service descriptions and mapping them into a logic-based knowledge base. In illustrative use cases involving simulated services for driving directions and flight information, the agent successfully produced customized plans. For instance, the same "Generic Travel Procedure" yielded different results for different users based on their individual DAML-encoded constraints (e.g., one user preferring to drive while another prefers to fly).

Limitations and Future Directions

While the logic-based approach is promising, several challenges remain. The authors acknowledge that while individual knowledge bases may be consistent, the Web as a whole is likely to contain conflicting information, posing a challenge for deductive reasoning. Furthermore, executing services with real-world consequences, such as financial transactions, requires robust security and transaction atomicity. Until the Semantic Web becomes ubiquitous, agents will still rely on information extraction wrappers for non-semantic sites. Finally, further research is needed to ensure that these deductive techniques can scale to the vast number of services available on the live Web.

Summary updated

BibTeX
@inproceedings{mcilraith2001mobilizing,
  author = {Sheila A. McIlraith and Tran Cao Son and Honglei Zeng},
  title = {{Mobilizing the Semantic Web with DAML-enabled Web Services}},
  booktitle = {SemWeb},
  year = {2001},
  url = {https://openreview.net/forum?id=jkwptxtKnD},
}
Figure