The VLDB Journal (2004) 13: 120–147 / Digital Object Identifier (DOI) 10.1007/s00778-004-0120-x ViSWeb – the Visual Semantic Web: unifying human and machine knowledge representations with Object-Process Methodology Dov Dori1,2 1 Technion, Israel Institute of Technology, Haifa 32000, Israel; e-mail: [email protected] 2 Massachusetts Institute of Technology, Cambridge, MA 02139, USA; e-mail: [email protected] Edited by V. Atluri. Received: December 14, 2002 / Accepted: November 28, 2003 Published online: February 6, 2004 – c Springer-Verlag 2004 Abstract. The Visual Semantic Web (ViSWeb) is a new Web is not a separate Web but an extension of the current one, paradigm for enhancing the current Semantic Web technol- in which information is given well-defined meaning, better en- ogy. Based on Object-Process Methodology (OPM), which abling computers and people to work in cooperation.” Indeed, enables modeling of systems in a single graphic and textual a major challenge in constructing a comprehensive Web-based model,ViSWeb provides for representation of knowledge over knowledge management system is the ability to reconcile the the Web in a unified way that caters to human perceptions apparent human-machine language orientation dilemma. On while also being machine processable. The advantages of the the one hand is the bulk of knowledge that has been published ViSWeb approach include equivalent graphic-text knowledge and continues to be gathered on the Web at an ever accelerat- representation, visual navigability, semantic sentence inter- ing rate. This knowledge is expressed in free natural language, pretation, specification of system dynamics, and complexity which in its raw form is currently indigestible to machines. management. Arguing against the claim that humans and ma- Indeed, as the Cyc project [11] has demonstrated, analysis of chines need to look at different knowledge representation for- unconstrained natural language by computers is far too diffi- mats, the principles and basics of various graphic and textual cult. On the other hand, current technologies that are aimed at knowledge representations are presented and examined as can- enabling Web-based knowledge management are developed didates for ViSWeb foundation. Since OPM is shown to be based on the premise that while humans prefer natural lan- most adequate for the task, ViSWeb is developed as an OPM- guage as the primary means of recording, communicating, and based layer on top of XML/RDF/OWL to express knowledge disseminating knowledge, machines must be fed with barely visually and in natural language. Both the graphic and the tex- human-intelligible, XML-like scripts, and if humans wish to tual representations are strictly equivalent. Being intuitive yet extract their meaning, they have to sweat to mentally compile formal, they are not only understandable to humans but are and decipher these scripts. also amenable to computer processing. The ability to use such Based on experience gained over the past half century bimodal knowledge representation is potentially a major step in developing artificial programming languages, the common forward in the evolution of the Semantic Web. wisdom has been that human-oriented and machine-oriented languages are necessarily mutually exclusive in terms of their Keywords: Semantic Web – Visual Semantic Web – Object- understandability: natural languages, which are intuitively un- Process Methodology – Conceptual graphs – Knowledge rep- derstood by humans, are much too difficult to be processed by resentation computers, while (programming) languages, which are easily “understood” by computers, require much training and effort to be deciphered by humans. The underlying assumption has been that the syntax of human-understandable language must necessarily be totally different than that of machine-digestible language. This view is clearly reflected by Berners-Lee and 1 The human-machine language orientation dilemma Hendler [4], who have noted that “. instead of asking ma- chines to understand people’s language, the new technology, A major assumption underlying the development of the Se- like the old, involves asking people to make some extra ef- mantic Web is that humans and machines must each use a dif- fort, in repayment for which they will get substantial new ferent format of knowledge representation. The first sentence functionality.” The ViSWeb approach was born as a result of the introduction of the Resource Description Framework of an attempt to provide the promised new functionality with- (RDF) [8] reads: “TheWorldWideWeb was originally built for out having to ask humans to make the extra effort required human consumption, and although everything on it is machine- by the current Semantic Web man-machine knowledge repre- readable, this data is not machine-understandable” (emphasis sentation dichotomy. Achieving this requires that people who in source). Berners-Lee et al. [5] have noted that “the Semantic invent these new environments make some extra effort in try- Dov Dori: ViSWeb – the visual semantic Web 121 ing to cater to human intuition while ensuring that computers this paper, which, to an outside human observer, superficially too will be able to process the same representation. suggests that they understand. The implicit, conventional wisdom assumption, accord- Synthesis of sentences that make perfect sense to humans ing to which human-readable and machine-readable formats using a subset of natural language while still being amenable are bound to be different, is the basis not only of RDF, but to reliable computer processing and compilation is not only also of OWL, the Web Ontology Language [32]. OWL is in- feasible, it has already been achieved to an astonishing level tended to describe the classes and relations between them of sophistication by Object-Process Language (OPL), a subset that are inherent in Web documents and applications. Ap- of English generated automatically on the fly in response to pendix B of [32] contains references to many Semantic Web graphic human input into an Object-Process Diagram (OPD). and ontology-related works. The introduction to OWL [32] OPL is the textual modality of Object-Process Methodology reads: “The World Wide Web, as it is currently constituted, re- (OPM) designed to favor the human over the machine in its sembles a poorly mapped geography....Inordertomapthis closeness to being natural yet formal to a sufficient degree to terrain more precisely, computational agents require machine- enable concise, unambiguous machine processing. readable descriptions of the content and capabilities of web The following sections contain a survey and assessment of accessible resources. These descriptions must be in addition various graphic and/or textual knowledge representation ap- to the human-readable versions of that information.” proaches and methods and expose the reader to principles of Challenging this assertion, according to which machine- Object-Process Methodology. OPM with the appropriate Vi- readable descriptions must be added on top of the human- sual Semantic Web extensions is then elaborated upon as a readable ones, this paper proposes a fresh approach to the issue human-understandable layer on top of RDF [8, 20] for speci- of knowledge representation over theWorldWideWebthrough fying knowledge over the Web. ViSWeb – the Visual Semantic Web. ViSWeb is founded on the premise that human and machine Web-based knowledge need not be represented by two distinct formats. The ViSWeb 2 Combining graphic (Visual Semantic Web) approach is based on Object-Process and textual knowledge representations Methodology (OPM) [14]. Using a bimodal representation of graphics and text, OPM models knowledge about systems of A powerful knowledge modeling and communication modal- various types and different complexity levels in a single model ity that is complementary to language is graphics. Diagrams that integrates structure and behavior. OPM, described in more are often invaluable for describing models of abstract things, detail below, combines a subset of natural language, called especially complex systems. The fact that people from the Object-Process Language (OPL), with a formal yet intuitive early caveman days to date have been using some kind of graphic model, a set of one or more Object-Process Diagrams sketching or diagramming technique to express their knowl- (OPDs) of exactly the same knowledge expressed in OPL. This edge or ideas is testimony to the viability of graphic repre- dual graphic-textual representation constitutes a solid founda- sentation. However, such representation of our knowledge is tion for generic knowledge representation over the Web. Since valuable only if it is backed by a comprehensive and consis- the ultimate objective of the Semantic Web is enhancing the tent modeling methodology. Such methodology is essential human ability to extract knowledge from the Web and under- if we want to represent knowledge, understand complex sys- stand it, it should involve human-understandable language. tems in any domain, and communicate our understanding to And since a combination of graphics and text is highly effec- others. An accepted diagramming method has the potential of tive as a knowledge modeling language [14], we follow this becoming a powerful modeling tool if it constitutes an unam- paradigm in designing ViSWeb, in which the human and ma- biguous language. In such visual formalism, each symbol must chine representations are effectively identical, enabling hu- bear defined semantics and