Like the fool I am and I’ll always be
I’ve got a dream
They can change their minds
But they can’t change me
Jim Croce, I’ve Got a Name (1973)

Is the search for intelligent agents the ultimate goal of the Semantic Web, or is it just a story to explain its potential? In any case, the idea of autonomous personal digital assistants exerts a strong attraction.

During three and a half years of research, I have been investigating how one day, autonomous pieces of software might use the Web similar to the way people can. This was inspired by Tim Berners-Lee’s vision of the Semantic Web [1], a layer on top of the existing Web that makes it interpretable for so-called intelligent agents. At one of the first conferences I attended, a keynote talk by Jim Hendler, co-author of the Semantic Web vision article, left me rather puzzled. Near the end of his talk—after convincing us all that the necessary technology is already out there—he posed the question: “so where are the agents?”

More than a decade of Semantic Web research unquestionably resulted in great progress, but nothing that resembles the envisioned intelligent agents is available. The Web has rapidly evolved, and many automated clients were created—yet all of them are preprogrammed for specific tasks. The holy grail of semantic technologies remains undiscovered, and researchers are sceptical as to whether it exists. The unbounded enthusiasm gradually makes place for pragmatism, as with any technology that advances on the hype cycle [3].

I had to maintain a realistic viewpoint during my search for solutions: trying to solve every possible challenge for autonomous agents would result in disappointment. The Semantic Web remains just a technology—albeit one that is assumed to make intelligent applications on the Web easier than its predecessors [2]. However, I believe the techniques discussed in this thesis advances the state of the art by making certain autonomous interactions possible that were significantly more difficult to achieve before. It cannot be a definitive answer to the quest for intelligent agents, but it might offer one of the stepping stones toward more autonomy for such agents.

Along the way, I will question some of the established principles and common practices on the Web. In particular, I will examine how we currently approach software building for the Web and plea for several changes that can make it more accessible for machines. As semantic technologies were never meant to be disruptive, the presented methods allow a gradual transition, backward-compatible with the existing Web infrastructure.

This thesis is structured in 8 chapters. After this introductory chapter, I will discuss the following topics:

• Chapter 2 – Hypermedia introduces the evolution of hypertext and hypermedia into the current Web. We detail how the REST architectural style has influenced the Web and examine why the Web’s current hypertext design is insufficient to support autonomous agents. This leads to the three research questions that drive this thesis.
• Chapter 3 – Semantics sketches the main components of Semantic Web technology and zooms in on the vision of intelligent agents. We discuss Linked Data as a pragmatic view on semantics. Finally, we elaborate on the relation between hypermedia and semantics on the Web.
• Chapter 4 – Functionality argues that machine clients need a way to predict the effect of actions on the Web. It introduces my work on RESTdesc, a lightweight approach to capture the functionality of Web APIs. a hypermedia-driven process for agents can offer an alternative to predetermined and rigid interaction patterns.
• Chapter 5 – Proof discusses my work on goal-driven Web API composition and the importance of proof and trust in the context of autonomous agents. We reconcile the error-prone act of API execution with the strictness of first-order logic and proofs.
• Chapter 6 – Affordance addresses an issue with the Web’s linking model: information publishers are responsible for link creation, yet they have insufficient knowledge to provide exactly those links a specific client needs. I introduce the concept of distributed affordance to generate the needed links in a personalized way.
• Chapter 7 – Serendipity questions the current way of Web application development. It proposes the use of semantic hypermedia as an enabling mechanism for applications that adapt to a specific client and problem context.
• Chapter 8 – Conclusion reviews the content of the preceding chapters, recapitulating the answers to the research questions that guide this thesis.

This thesis has been conceived as a book with a narrative, preferring natural language over mathematical rigorousness to the extent possible and appropriate. Since no act of research ever happens in isolation, I will use the authorial “we” throughout the text, except in places where I want to emphasize my own viewpoint.

The underlying motivation is to make this work more accessible, while references to my publications guide the reader towards in-depth explanations.

I hope this introduction may be the start of a fascinating journey through hypermedia and semantics. I learned a lot while conducting this research;
may the topics in this book inspire you in turn.

References

[1]

Tim Berners-Lee, James Hendler, and Ora Lassila. The Semantic Web. Scientific American, 284(5): 34–43, May 2001.

[2]

Christian Bizer, Tom Heath, and Tim Berners-Lee. Linked Data – the story so far. International Journal on Semantic Web and Information Systems, 5(3): 1–22, March 2009.

[3]

Jackie Fenn and Mark Raskino. Understanding Gartner’s hype cycles. Technical report. 2 July 2013.

Comments and questions