Conclusions

At the start of this book we outlined the challenges of automatic computer based processing of information on the Web. These numerous challenges are generally referred to as the ‘vision’ of the Semantic Web. From the outset, we have attempted to take a realistic and pragmatic view of this vision. Our opinion is that the vision may never be fully realized, but that it is a useful goal on which to focus. Each step towards the vision has provided new insights on classical problems in knowledge representation, MASs, and Web-based techniques. Thus, we are presently in a significantly better position as a result of these efforts. It is sometimes difficult to see the purpose of the Semantic Web vision behind all of the different technologies and acronyms. However, the fundamental purpose of the Semantic Web is essentially large scale and automated data integration. The Semantic Web is not just about providing a more intelligent kind of Web search, but also about taking the results of these searches and combining them in interesting and useful ways. As stated in Chapter 1, the possible applications for the Semantic Web include: automated data mining, e-science experiments, e-learning systems, personalized newspapers and journals, and intelligent devices. The current state of progress towards the Semantic Web vision is summarized in Figure 8.1. This figure shows a pyramid with the human-centric Web at the bottom, sometimes termed the Syntactic Web, and the envisioned Semantic Web at the top. Throughout this book, we have been moving upwards on this pyramid, and it should be clear that a great deal of progress that has been made towards the goal. This progress is indicated by the various stages of the pyramid, which can be summarized as follows: • The lowest stage on the pyramid is the basic Web that should be familiar to everyone. This Web of information is human-centric and contains very little automation. Nonetheless, the Web provides the basic protocols and technologies on which the Semantic Web is founded. Furthermore, the information which is represented on the Web will ultimately be the source of knowledge for the Semantic Web.

Reactive agents

In the previous chapter we described three languages for representing knowledge on the Semantic Web: RDF, RDFS, and OWL. These languages enable us to create Web-based knowledge in a standard manner with a common semantics. We now turn our attention to the techniques that can utilize this knowledge in an automated manner. These techniques are fundamental to the construction of the Semantic Web, as without automation we do not gain any real benefit over the current Web. There are currently two views of the Semantic Web that have implications for the kind of automation that we can hope to achieve: 1. An expert system with a distributed knowledge base. 2. A society of agents that solve complex knowledge-based tasks. In the first view, the Semantic Web is essentially treated a single-user application that reasons about some Web-based knowledge. For example, a service that queries the knowledge to answer specific questions. This is a perfectly acceptable view, and its realization is significantly challenging. However, in this book we primarily subscribe to the second view. In this more-generalized view, the knowledge is not treated as a single body, and it is not necessary to obtain a global view of the knowledge. Instead, the knowledge is exchanged and manipulated in a peer-to-peer (P2P) manner between different entities. These entities act on behalf of human users, and require only enough knowledge to perform the task to which they are assigned. The use of entities to solve complex problems on the Web is captured by the notion of an agent. In human terms, an agent is an intermediary who makes a complex organization externally accessible. For example, a travel agent simplifies the problem of booking a holiday. This concept of simplifying the interface to a complex framework is a key goal of the Semantic Web. We would like to make it straightforward for a human to interact with a wide variety of disparate sources of knowledge without becoming mired in the details. To accomplish this, we want to define software agents that act with similar characteristics to human agents.

Web knowledge

In the introductory chapter of this book, we discussed the means by which knowledge can be made available on the Web. That is, the representation of the knowledge in a form by which it can be automatically processed by a computer. To recap, we identified two essential steps that were deemed necessary to achieve this task: 1. We discussed the need to agree on a suitable structure for the knowledge that we wish to represent. This is achieved through the construction of a semantic network, which defines the main concepts of the knowledge, and the relationships between these concepts. We presented an example network that contained the main concepts to differentiate between kinds of cameras. Our network is a conceptualization, or an abstract view of a small part of the world. A conceptualization is defined formally in an ontology, which is in essence a vocabulary for knowledge representation. 2. We discussed the construction of a knowledge base, which is a store of knowledge about a domain in machine-processable form; essentially a database of knowledge. A knowledge base is constructed through the classification of a body of information according to an ontology. The result will be a store of facts and rules that describe the domain. Our example described the classification of different camera features to form a knowledge base. The knowledge base is expressed formally in the language of the ontology over which it is defined. In this chapter we elaborate on these two steps to show how we can define ontologies and knowledge bases specifically for the Web. This will enable us to construct Semantic Web applications that make use of this knowledge. The chapter is devoted to a detailed explanation of the syntax and pragmatics of the RDF, RDFS, and OWL Semantic Web standards. The resource description framework (RDF) is an established standard for knowledge representation on the Web. Taken together with the associated RDF Schema (RDFS) standard, we have a language for representing simple ontologies and knowledge bases on the Web.

Agent communication

In the preceding chapters, we have described in detail the various techniques that can be used to construct reasoning agents for the Semantic Web. However, as we have repeatedly emphasized, this reasoning capability provides only part of the functionality that will be instrumental in the construction of Semantic Web applications. To supply the remaining functionality, we must also provide a communicative capability to the agents that we construct. This capability will allow our agents to interact and cooperate with other agents on the Semantic Web, and thereby realize more complex tasks than could be accomplished with a single agent in isolation. The focus of this chapter is on how we can design and build agents which can interact together successfully on the Semantic Web. To address the issues of communication and coordination in this environment, it is useful to take our inspiration from human social systems, where these issues are solved in the real world. Following this approach, we treat a group of agents as a society, where these agents typically share a common interest, e.g. solving a specific type of problem. To become a member of this society, an agent must agree to observe certain rules and conventions. In return, the agent can itself benefit from the society, e.g. the expertise of other agents. In this societal view, it becomes possible to define conventions for the agents to follow, and the incentives for agents to operate together are made clear. This in turn makes the challenges of agent-based communication and coordination more manageable. The societal view is a popular approach in the field of MASs, and has important consequences for the way we define our systems. Our discussion is structured around the following five considerations, which must be addressed if we are to design agents that can successfully participate in a society: 1. The ability to communicate with other agents. 2. A basis on which to understand what is being communicated. 3. The ability to structure communication into coherent patterns.

Semantic Web services

In this book we have been consistently directed by the vision of the Semantic Web. This vision can be summarized as the ability for computers to automatically use information on the Web in a similar way to humans. In particular, we want to be able to retrieve, comprehend, and exchange knowledge using automated techniques. At this point we have defined all of the main techniques that can be used to realize these goals. A summary of the four key techniques that we now have at our disposal is presented below: 1. We have the ability to represent knowledge in a form suitable for automated processing. This ability is provided by the definition of ontologies, which provide structure to knowledge. 2. We can construct entities, called agents, which act on behalf of humans and solve specific goals. We have presented many different techniques that can be used to construct these agents, dependent on the purpose for that the agents to be applied. 3. We can reason about the knowledge that we represent to answer specific questions. This can be accomplished by query answering techniques, or by complex inferences over the knowledge, guided by the ontology. 4. Our agents can communicate with other agents, and form societies based on common interests. Within these societies, agents can collaborate towards the resolution of common goals, which could not be accomplished by individual agents alone. The purpose of this penultimate chapter is to show how we can harness and combine these four key techniques to build systems and applications for the Semantic Web. As stated in Chapter 1, Semantic Web applications are not constructed statically in the traditional manner. Instead, these applications are constructed dynamically, at run-time, from combinations of services, termed knowledge services. Our presentation is designed to answer the two key questions below: 1. How can we construct knowledge services that encompass the various capabilities that we have available? 2. How do we compose knowledge services into applications that can accomplish specific tasks?

The Semantic Web

At the present time, the Web is primarily designed for human consumption and not for computer consumption. This may seem like an unusual state of affairs, given that the web is vast and mature computerized information resource. However, we must recognize that the computer is presently used as the carrier of this information, and not as the consumer of the information. As a result, a great deal of the potential of the Web has yet to be realized. This book explores the challenges of automatic computer-based processing of information on the Web. In effect, we want to enable computers to use Web-based information in much the same way as humans presently do. Our motivation is that computers have a brute-force advantage over humans. Where we can gather and process information from a handful of Web-based sources, a computer may download and compare thousands of such sources in a matter of seconds. Nonetheless, despite the apparent simplicity of this task, there are a great many issues that must be addressed if we are to make effective use of this information. As a result, the automated processing of Web-based information is still in its infancy. In this book, we show how many different techniques can be used together to address this task. The automated processing of information on the Web is principally an issue of scale. There are many existing techniques in Computer Science and Artificial Intelligence (AI) that may be appropriate to the task. However, there are significant issues that must be addressed if we are to scale up these techniques for use on the Web. Therefore, we present a detailed overview of the current state of the art, with a particular emphasis on practical solutions. The methods and technologies that we present in this book are of importance to all computer practitioners, as they will shape the future evolution of the Web. To appreciate the challenges of computer-based consumption of Web-based information, we consider the following scenario. Suppose we are searching the Web for information on a specific ailment.

Reasoning on the Web

The techniques that we can use to construct rational agents were presented in Chapters 3 and 4. In these chapters, we identified a variety of general purpose reasoning techniques that can be used by agents to accomplish goal-directed behaviour. In this chapter, we turn our attention to the construction of agent-based reasoning processes that are directed towards the Semantic Web. These processes are based on the general techniques that we described previously, but are tailored specifically for the Semantic Web. In particular, our reasoning processes are designed to operate directly on Semantic Web knowledge expressed in RDF, RDFS, and OWL documents. These reasoning processes are essentially specialized kinds of deductive reasoning systems. It is important to appreciate that the reasoning processes that we describe in this chapter have certain limitations. For each technique, we will only be able to solve certain classes of problems. These limitations are a direct consequence of the representation that we use. Thus, it is necessary that we understand what kinds of problems we can solve in each approach, as this will determine the kinds of Semantic Web applications that we can construct. In effect, we are seeking to answer two key questions: 1. What kinds of reasoning can we perform with our knowledge? 2. How do we specify the problems that we wish to solve? The first question concerns the representation of the knowledge. In Chapter 2 we stated that there is a trade-off between expressibility and efficient reasoning. In general, the more features that we have in the representation language, the more difficult it is to reason with the language. The second question concerns the definition of the reasoning process itself. As we show in this chapter, there is a further trade-off between specification complexity and reasoning power. In general, the more complex our specification formalism, the more difficult it is to reason efficiently. In this chapter, we discuss two main approaches that we can use to define reasoning processes for the Semantic Web: query languages and logic-based formalisms.

Practical reasoning and deductive agents

In constructing a reactive agent system, we explicitly define the behaviour of each agent. This behaviour is predefined, and dependent on events in the environment. We now consider a more powerful kind of agent that can make decisions on its own, i.e. an agent with proactive behaviour. Our motivation is the construction of agents with capabilities that are closer to the way that we reason as human beings. Our starting point in this approach is to base the internal processes of the agent directly on current understanding of how human reasoning is performed. This is the principle behind the design of a practical reasoning agent. Practical human reasoning is directed towards actions, that is, figuring out what to do. This is different from purely logical reasoning, which is directed towards beliefs. Human reasoning is believed to consist of two distinct phases: 1. The first phase is deliberation, in which we decide what state of affairs to achieve. 2. The second phase is means–ends reasoning, in which we decide how to achieve the desired state of affairs. To better illustrate human reasoning, it is helpful to consider a small example. Suppose that I wish to find a method of transportation in order to get to work each day. I would typically proceed by considering the various available options and weighing up the pros and cons. For example, I may consider travelling by car, but the available parking may be insufficient. This process of decision-making is deliberation. Once I have fixed upon an appropriate method of transport, e.g. by bicycle, then I must decide how to bring about a situation where this method of transport is possible. This process is means–ends reasoning, and the result is a plan of action. For example, my plan may involve: obtaining the money for the bicycle, finding a shop that sells an appropriate bicycle, and then purchasing the bicycle. If I am able to successfully follow the plan, then I will have reached the intended state of affairs, i.e. I will be able to travel to work by bicycle.