address some conceptual shortcomings of the semantic and pragmatic web ... will create the network of documents whose content could be automatically processed in a much more elastic .... Diagrammatic methods are also easy to use. 6.
THE SEMANTIC + PRAGMATIC WEB = THE SEMIOTIC WEB Ahti-Veikko Pietarinen Department of Philosophy University of Helsinki P.O. Box 9, FIN-00014 University of Helsinki
ABSTRACT The desideratum for the concept of a semiotic web is proposed. It combines both the recent semantic web and the pragmatic web initiatives but surpasses them in being faithful to Charles S. Peirce’s pragmatic approach to inquiry. I will address some conceptual shortcomings of the semantic and pragmatic web proposals, and suggest that in a semiotic web, one can to operationalise its basic ideas in an effective way. Accordingly, a foundational approach to multi-agent systems is proposed that goes beyond both semantic and pragmatic notions by employing a logic of questioning and answering, and by taking agents as roles in a dialogical semiotic inquiry of all signs mediated in the universe of the web. The outcome is that in the web realm, semantic and pragmatic elements cannot and should not be separated from each other, and that their merger has significant emergent properties matching Peirce’s sign theory. KEYWORDS Semantic web, pragmatic web, semiotic web, web philosophy, multi-agent systems, Charles S. Peirce.
1. INTRODUCTION Both semantic and pragmatic web visions share their concern for the inadequacy of the current conceptual scheme of the web. Semantic web aims at providing a “logical analysis” of the data, while pragmatic web adds the human perspective to it. Their conceptual underpinnings have not been sufficiently addressed, however, and the focus of discussion has bogged down to syntactic properties of language technologies. The purpose of this paper is to argue, by means of a conceptual analysis, that both of these approaches are somewhat inadequate, but that their merger, when conceived from the perspective of the semiotic and logical approach to inquiry, has significant emergent features. It is thus a promising framework for the concept of a web for which both semantic and pragmatic web initiatives seem inadequate alone.
2. THE ACHILLES’ HEEL OF A SEMANTIC WEB 2.1 A semantic web: meaning from information People have always strived to give meaning to information, and today the reason for this is the desire to make humans and computers interoperable. Accordingly, the goal of the next-generation web technology is to define meaning into web documents so that it can be processed and understood by machines [1]. The increasingly popular albeit not yet very widely implemented approach known as a semantic web typically uses axiomatic mark-up methods that instead of plain keywords, define the class and subclass hierarchies and the relations between the concepts that appear on the page [2,3,10]. This metadata information provides the main component of what the “semantics” or “meaning” of the document is supposed to be. According to this project, it is hoped that the increased production of metadata and ontology languages will create the network of documents whose content could be automatically processed in a much more elastic and adaptable manner than in the standard syntax-driven string-matching and link-weighted search methods.
981
International Conference WWW/Internet 2003
2.2 This is Peirce’s semantics, but not it is not meaning On the face of it, one thinks that this approach has nothing to do with semantics as a comprehensive theory of meaning. However, it in fact is the practical present-day account of a Peircean notion of semantics, because it provides meaning of the data or a code by translation [7]. This is because for Peirce, semantics is, in fact, a theory of translation, a rendition of a given symbolic statement into some other statement, diction, or paraphrase, or into some other language or perhaps a dictionary-like definition of it. But to get from this conception to what the meaning of the data really is represents a considerably more complicated challenge. To get meaning, one needs to take into account the effects that any figment of data may have on its interpreter. This problem was addressed in Peirce’s sign theory more than a century ago, but there are precious little signs that it has been acknowledged in current programmes advancing the narrowlyfocussed concept of “semantics” for the web.
3. THE ACHILLES’ HEEL OF PRAGMATIC WEB 3.1 Wanted: the utterer and the interpreter First and foremost, this narrow understanding of semantics lacks the semiotic components of the utterers and the interpreters of the data, indispensable in pragmatic approaches to inquiry. And so, one still needs to understand how the metadata, such as one provided by the schema of Resource Description Framework (RDFS) or its ilk, will be connected to the interpreters and objects of data. This connection defines the pragmatic meaning of data.
3.2 The vision of a pragmatic web However, as such this does not seem to hover at the back of the vision of a pragmatic web [6]. The perspective to the pragmatic web is rather in bringing the community of inquirers, most notably web users, to bear on the issues of the purpose of information. While such intentions and contexts of users surely play a significant role in pragmatic accounts of meaning, and while these researchers are certainly right in criticising the semantic web approach in its limitation to the metadata idea ignoring the communities of human users and engineers, this approach sidesteps the perhaps more profitable possibility of incorporating truly semiotic pragmatics into the automatised and computational level of the web. It correctly notes the insufficiency of the vision of the semantic web that does not think that all human users, not just knowledge engineers in this self-critical approach are indispensable. It asserts that new meanings or concepts do not simply emerge by adding more and more structural features onto the web pages and by linking them more and more efficiently.
3.3 Towards semiotic pragmatics Even so, there is still a need for methodeutic, or semiotic pragmatics, that makes contact with the third main aspect of inquiry beyond grammar and logical semantics, but taking place solely between computerised agents. This component addresses issues in pragmatic understanding of natural and artificial languages.
4. SEMIOTISING THE AGENT SYSTEMS 4.1 Dialogical semiosis From the perspective of multi-agent systems, precisely this pragmatic web challenge may be brought into life. The challenge involves an attempt of building agents, or pieces of advanced software, which are designed to
982
THE SEMANTIC + PRAGMATIC WEB = THE SEMIOTIC WEB
play the different semiotic roles of the quasi-utterers and the quasi-interpreters (“quasi” because they may not be connected with actual minds, i.e., are applicable for artificial systems and artefacts). That is, they would play the different positions in the cycles of dialogical semiosis as prescribed by Peirce’s theory. This is the way in which they are intended to contribute to the generation of new objects and the evolution of new meanings in the web. Ultimately, this has been for a long one of the few truly respected routes towards genuinely functioning and conceptually appealing systems of artificial intelligence.
4.2 Conceptual challenges for agent systems Agent systems still lack the truly goal-directed specifications of processes (but see [4]). Only when that is accomplished, they could be seen to create habits and produce wherewithal to revise them. This is a long way off. Autonomous, proactive, reactive and social agents need to build second-order evaluations of their own strategies, noting when “habit-changes” occur, namely when logical interpretants are produced in the quasiminds of agents as the end products of the process of semiosis that terminates or is about to terminate. They need to learn whenever signs signify meaningful logical information to one of the individual agents. But to know when that is to happen depends on a correct evaluation of those habits that already are in agents’ possession.
5. QUESTIONING THE WEB 5.1 Knowledge web Elements of such goal-directed agent systems are emerging in the vision of a knowledge web [5]. It tries to overcome these shortcomings and supersede both the semantic and the pragmatic web enterprises by taking agents as constructors building a huge question-answering system on the web data, and responding queries on an information-need basis. This is certainly a long way off, because one needs to accomplish two things: one needs to (i) have a comprehensive logic of questions and answers, and to (ii) define a workable possibleworlds structure of the web. Neither of these tasks has been accomplished as yet.
5.2 Towards a comprehensive logic of questions and answers New web visions facilitate self-supporting discovery of new information. To do this, in relation to the item (i) above, agents need to ask questions from the oracle (the web). This importance of questioning has not been sufficiently noted. The central quest in accomplishing automatised question answering in information discovery is to find out logical relations between questions and answers. This, in turn, hinges on the theories of presuppositions. (The task is also related to the extensions of Peirce’s theory of abduction.) This is because requests for information can be viewed as epistemic statements. The query “Bring it about so that S” (S is any proposition of one’s preferred modelling language) has a meaning in non-imperative epistemic sentence “I know that S”. As to the second item, the web nodes are to be viewed as knowledge providers, and via that emerging structure the epistemic statements can be translated to mean that “the user knows S in the information state w if and only if S holds in all the web nodes accessible from w”.
5.3 Model-checking the web There then needs to be agreed initial states for the users of information. But as they are software agents querying other software agents, w codifies the knowledge in the position they have reached within a modelchecking game on the space of the relational structure of web nodes. The structure that these agents create in the course of their question-answering procedures is a possible-worlds structure, exhibiting the degree of connectivity between web pages.
983
International Conference WWW/Internet 2003
5.4 Conceptual graphs An alternative, complementary way to the building of a knowledge web uses conceptual graphs [5,11]. They are descendants of Peirce’s existential (beta) graphs in modern clothing [8,9,11]. Their purpose in the domain of a knowledge web is to provide high-quality representational formalisms to describe workings of software agents. Diagrammatic methods are also easy to use.
6. TOWARDS A SEMIOTIC WEB We need both semantic and pragmatic web conceptions. But due to their disjoint initiatives, they cannot be put together without considering both from a birds-eye perspective. Our proposal is that, briefly, the semantic and the pragmatic web enterprises give rise to the conception of a semiotic web that no longer possesses the weak concept of semantics in the extant semantic web proposals, and which operationalises the pragmatic web conception by viewing agents in multi-agent systems where agents can play the semiotic roles of the utterers and the interpreters. Among other things, this is one of the few ways in which pragmatic aspects of language, including its assertoric, demonstrative and contextual use in formal dialogue and negotiation systems can truly be addressed. The adjoined immense repertoire provided by Peirce’s sign theory is virtually inexhaustible in its resources by which to increase the level of interoperability between any entities that mediate signs, be they actual end users or computerised agents. Its technological realisation in terms of a semantic web shows what the communities of computerised inquiry consists of, and how humans interacting with software actually create their triadic forms of representations between the signs, the data, and the interpretants. Finally, it remains to be seen whether software agents will give rise to interpretants of the web data in an equally satisfactory manner.
ACKNOWLEDGMENTS The work has been supported by the Academy of Finland (“Communications in the 21st Century: The Relevance of C.S. Peirce”, Project No: 104262), and by the Georg and Ella Ehrnrooth Foundation.
REFERENCES [1] Berners-Lee, Tim, 1999. Weaving the Web: The Original Design and Ultimate Destiny of the World Wide Web by its Inventor. Orion Business Books, London.
[2] Berners-Lee, Tim and Eric Miller, 2002. The semantic web lifts off, ERCIM News 51, pp. 9–11, http://www.ercim.org/publication/Ercim_News/enw51/berners-lee.html.
[3] Fensel, Dieter and Mark A. Musen, eds., 2001. The Semantic Web: A brain for humankind, IEEE Intelligent Systems 16, pp. 24–25.
[4] Kraus, Sarit, 2001. Strategic Negotiation in Multiagent Environments, MIT Press, Cambridge, Mass. [5] Mineau, Guy W., 2002. A first step toward the Knowledge Web: Interoperability issues among conceptual graph
based software agents, part I”, in Uta Priss et al., eds., Conceptual Structures: Integration and Interfaces, Lecture Notes in Computer Science 2393, Springer, pp. 250–260. [6] de Moor, Aldo et al., 2002. Towards a pragmatic web, in Uta Priss et al., eds., Conceptual Structures: Integration and Interfaces, Lecture Notes in Computer Science 2393, Springer, pp. 235–249. [7] Peirce, Charles S., 1980–. The Writings of Charles S. Peirce. 6 vols. Vol.1, Max Fisch et at., vol. 2, ed. by Edward C. Moore et al., vols. 3-5, ed. by Christian Kloesel et al., vol. 6, ed. by the Peirce Edition Project, Indiana University Press, Bloomington. [8] Pietarinen, Ahti-Veikko, 2003. Peirce’s game-theoretic ideas in logic, Semiotica 144, 2003, pp. 33–47. [9] Pietarinen, Ahti-Veikko, 2003. Diagrammatic logic and game-playing, in Malcolm, Grant, ed., Multidisciplinary Approaches to Visual Representations and Interpretations, Elsevier Science. [10] The Semantic Web Agreement Group, 2002. What is the Semantic Web?, http://swag.webns.net/. [11] Sowa, John, 1998. Conceptual graphs, in P. Bernus et al., eds., Handbook on Architectures of Information Systems, Springer, Berlin, pp. 287–311.
984