INTRODUCTION: FROM PRESENTATION-LEVEL INTEGRATION TO CONCEPT-LEVEL INTEGRATION IN PORTALS
In general, portals are regarded as gateways to networked information and services, facilitating access to other related sites. Typically, portals provide transparent one-stop access to functionalities needed in a common context and make these appear as a single integrated application. Although these functionalities are implemented and made available by heterogeneous applications, they are integrated by presenting the output of such networked applications side-by-side, with only limited interaction between them. This superficial integration on the presentation level is quite useful, but it has considerable drawbacks compared to basing portals on the notion of subjects to achieve “seamless knowledge” and the “semantic superhighway” (Pepper, 2004, 2006; Pepper & Garshol, 2004). Consider these challenges:
• Knowledge Organization: How can a portal user be directed in a principled way from a given asset about a subject to other relevant assets about this subject, to other related subjects and the assets linked to them?
• Portal Integration: How can all assets about a given subject be made virtually accessible from one page in the portal, even if the content about the subject is distributed to several independently maintained portals?
Both challenges require an appropriate information architecture with an integration on the conceptual level. Content- and semantics-based portal approaches allow expressing a domain knowledge model (concepts and their interrelations) and connecting it to related resources.
Afterbriefly sketching how topic-oriented portals (TOPs) are related to knowledge portals, ontology-based and semantic portals, we discuss Topic Maps-based portals (TMPs), a specific form of topic-oriented (or subject-oriented) portals with advantages in their creation, usage and maintenance. What are examples for such portals and their virtual integration? Which elements are appropriate for their information architecture? How could a subject-oriented information architecture be based on knowledge organization systems (KOS)?
For the remainder of this article, it is assumed that the reader is acquainted with basic Semantic Web concepts, including Topic Maps (Ahmed & Moore, 2005; Passin, 2004).
BACKGROUND: TOPIC-ORIENTED PORTALS
A topic-oriented (or subject-oriented) portal1 is a Web-based information or knowledge portal application where informational content is presented in relationship to subjects and whose structure, or “information architecture” (Rosenfeld & Morville, 2002), makes use of Semantic Web technologies. A TOP is not necessarily a portal dedicated to a special narrow selection of topics or subjects. A TOP typically presents a page-per-subject view of a knowledge domain: Each subject has its own “homepage” that displays the information related to that subject and the relationships between the subject and other subjects in the knowledge domain. The pages about the related subjects may reside on the same portal, or on separate portals. A shallow ontology, specifying the subjects of interest and their interconnecting relationships, is the basis of the information architecture which structures the navigation, content integration and rendering, and the search facilities of content-rich Web sites. This knowledge structure on the ontology layer is by design separated from the resource layer. The resource layer describes which resources (assets such as documents) exist, and the connection between both layers specifies which resources are relevant to which subjects, and in which way. By this separation of layers, the knowledge structure becomes portable. This means that it can be superimposed upon different content, thus grouping assertions and relevant resources referring to the same subject together, and identifying and showing assertions and resources referring to related subjects in a principled way. A TOP can be implemented with different Semantic Web technologies, in particular with RDF/OWL or Topic Maps.
Related Work on Knowledge Portals, Ontology-Based and Semantic Portals
No reference to TOPs is made in recent work on knowledge, ontology-based and semantic portals (cf. Hadrich & Priebe, 2005a, b; Lara et al., 2004; Lausen et al., 2005), or (Hartmann & Sure, 2004) for the SEAL (SEmantic portAL) conceptual framework. However, the approaches are all closely related:
A knowledge portal is an information portal supporting knowledge workers in their tasks. It is a specific type of enterprise portal, comprising support for information content storage and retrieval, organizational communication and group collaboration (cf. Detlor, 2004, p. 13). An ontology-based portal (cf. Staab et al., 2000) is a portal employing ontologies as its semantic backbone, mainly for information integration, navigation and search. A semantic portal is a portal using Semantic Web technologies.
A TOP can be a knowledge portal, since the subject-centric integration supports knowledge management and organization, and knowledge workers may be supported in all three of Detlor’s dimensions. Since the knowledge net has an ontologic layer, it is an ontology-based portal, and it is a semantic portal, because Semantic Web technologies are used for the representation and manipulation of the ontology and the metadata about the resources. The creation of semantically linked Web pages from Semantic Web content can result in TOPs (Hyvonen, Valo, Viljanen, & Holi, 2003). These authors acknowledge the similarity of their approach to Topic Maps, except they infer the linkage structure instead of specifying it.
topic maps-based portals
This article focuses on Topic Maps-based (or Topic Maps-driven) portals (TMPs), that is, TOPs realized with Topic Maps. TMPs are “the most common application of Topic Maps today,” and “also by far the most visible” (Garshol, 2006c). Topic Maps are understood as standardized in the second edition of ISO/IEC 13250 Topic Maps (ISO 13250), with TMDM (Topic Maps Data Model) (ISO 13250-2) to become part of this standard. The interested reader is directed to the introductory book on XML Topic Maps (Park & Hunting, 2002) and to the discussion of research issues at the TMRA conferences (Maicher & Park, 2006).
Examples of Topic Maps-Based Portals
Almost any Topic Map rendered in a subject-centric way as interlinked Web pages, aggregating everything known about a particular subject, will lead to a TMP, for example, the so-called “Italian Opera Topic Map” by Steve Pepper.2
Every topic has its own “homepage” which displays the semantic relation of this topic to other topics and the assets (resources) relevant to this topic, and all pages are interlinked. Several TMPs are in practical use, predominantly in Norway, for example Kulturnett,3 a Norwegian public sector portal to cultural information. The IRS Topic Map4 by Michel Biezunski is a prominent example in English language. Barta (2004) communicates his experiences in developing a Perl-based knowledge portal using Topic Maps; Pepper and Garshol (2002) based theirs on building a TMP of conference papers. For more practical examples of TMPs in the public sector and their virtual integration, see Pepper (2004) and Garshol (2006c).
Examples of the virtual Integration of Topic Maps-based Portals
Consider you want to connect at least two portals such that topic pages about the same topic provided by both portals are virtually integrated (cf. the simple portal connecting scenario of Pepper, 2004). Portal A can ask portal B if it knows anything about this topic, update its own knowledge base with the answer on-the-fly, and present its changed pages to the user. For example, the three following Norwegian TMPs (of the Research Council Web site for young adults,5 the public site of the Consumer Association,6 and the biosecurity portal of the Department of Agriculture7) have topics about the common subject “genetically modified food” which they can mutually share, based on published subjects. Connecting portals is just one of four use cases for TMRAP, a remote access protocol for Topic Maps (Garshol, 2006a). Using published subjects and TMRAP as a vehicle to realize the idea of “seamless knowledge,” portals can also share little Topic Map fragments, thus automatically syndicating, synchronizing and aggregating knowledge structures and the accompanying resources (Garshol, 2006a, c).
Elements of the Information Architecture of Topic Maps-Based Portals
Topic Maps and TOPs match well, because Topic Maps already exhibit all features for the implementation of TOPs. In particular, basic principles of Topic Maps and knowledge organization aid in the design of an appropriate information architecture:
• Subject Centric View: In contrast to the more resource-centric RDF, with subjects and topics, Topic Maps are by design subject-centric, making it easier to talk about subjects.
• Semantic Interoperability with RDF: All work on RDF-based portals can be reused, since Topic Maps are semantically interoperable with RDF (Garshol, 2003, 2005; Garshol & Naito, 2004; Ontopia, 2003; Pepper, Presutti, Vitali, Garshol, & Gessa, 2006; Pepper, Vitali, Garshol, Gessa, & Presutti, 2005). Because Topic Maps explicitly disclose additional information (Garshol, 2006b), a mapping from RDF to Topic Maps is an up-conversion (Garshol, 2003, 2005).
• Constructs on the Ontology and the Resource Layer: In contrast to RDF, Topic Maps discern between the ontology and the resource layer. Constructs are provided for defining assertions (expressed as topics and associations) (ontology layer), and for referencing relevant content (resource layer). An occurrence represents the relationship between a subject and an information resource, and the occurrence type describes the nature of the relationship between the subjects and information resources linked by the occurrence of that type (ISO 13250-2). Thus one can systematically and flexibly specify which resources related to a given resource shall be shown in a TMP. One only needs to define both the occurrence relations from one resource to the ontology and back from the ontology to another resource, and the association types connecting two topics within the ontology.
• Identity Management and Merging Capabilities for Cross-Portal Collaboration and Aggregation: As a cure to the Web’s identity crisis, with subject indicators Topic Maps provide a mechanism for unambiguously identifying and addressing subjects, also applicable with RDF (Maicher, 2004; Pepper & Schwab, 2003). Published subjects (PSIs) enable a semantically interoperable information architecture (OASIS Published Subjects TC, 2003; Pepper, 2006). According to SLUO (the Subject Location Uniqueness Objective) or collocation objective of Topic Maps, everything known about a subject should be accessible from one (virtual) place, thus for each unique subject there should be only one proxy (an addressable representation in the computer) (Newcomb, Hunting, Algermissen, & Durusau, 2003, par. 2.26). Topics serve as points of collocation, also between federated portals. Technically, Topic Maps have features for merging topics, for example, based on subject indicators.
• Loose-Coupled Information Architecture, Ontologies, and Data Hubs: Topic Maps combine well with all these concepts. In the case of a TMP, the underlying Topic Maps specify a shallow ontology, the system of types of topics, associations, and occurrences that together define the classes of things and relationships between things (Ahmed & Moore, 2005), “a model for describing the world that consists of a set of types, properties, and relationship types” (Garshol, 2004). “[A] Topic Map can serve as the data hub of a loose-coupled information architecture, allowing new data sources to be added and merged with the content and other data sources that drive the site.” See Networked-Planet (2005) for advantages of this approach.
• Knowledge Organization: Topic Maps also combine naturally with principles of knowledge organization (Sigel, 2002). It is recommended that organizations planning to create sophisticated TMPs build their information architecture on one or several KOS. They should leverage simple subject structures, traditional taxonomies, thesauri, or faceted classification schemes to shallow ontologies grounded in published subjects (Pepper, 2006). One approach could be to express these KOS using the SKOS (Simple Knowledge Organisation System) (Miles & Brickley, 2005; Miles et al., 2005) RDF vocabulary (for an example see SWED, the semantic Web environmental directory8), up-converting from RDF to Topic Maps.
• Feasible, Lightweight, Yet Extensible Ontologies: In portal practice, ontologies described with Topic Maps are rather applied than ontologies represented with OWL, because they are easier to design and use for end users and information architects. In addition, with Topic Maps one can start with a lightweight but quite expressive approach and extend the semantics as needed.
• Topic Map Design Patterns: Ready-made Topic Map design patterns can be applied for information architecture (Ahmed, 2003b).
• Portal Federation: Portals can “talk to each other,” exchanging Topic Map fragments, for example, with TMRAP, an abstract Web service interface for remote access to Topic Maps (Garshol, 2006a; Pepper & Garshol, 2004), or TMIP (Barta, 2005). This exchange allows several federated portals to cooperate P2P-like (Ahmed, 2003a), which will support decentralized and distributed knowledge management and emergent knowledge structures (Bonifacio, 2006).
FUTURE TRENDS
We expect more organizations specifying their ontologies as Topic Maps, and more portals being topic- or subj ect-oriented and using published subjects. The number of TMPs based on knowledge organization principles and SKOS will increase, working towards the collocation principle (everything known about a given subject shall be virtually accessible from one place). More semantic portals will be virtually coupled, exchanging Topic Map fragments via Web services, integrating knowledge in “knowledge hubs” (Ahmed & Moore, 2006; Garshol, 2006a), and enhancing decentralized knowledge management and emergent knowledge organization p2p-like. Semantic interoperability between such applications, but also between Topic Maps and RDF will likely increase. A tighter intertwining of the research and application of knowledge portals, ontology-based and semantic portals with TOPs is recommended.
CONCLUSION
Most existing portals integrate information from various sources only on the presentation level, not on the concept level, and this is quite useful on its own. However, in order to present resources relevant to other resources in a principled way, or to be able to interconnect resources about the same subject from several portals to a virtual portal, a topic- or subject-oriented information architecture is necessary. Although TOPs can be implemented with different Semantic Web technologies, basing their information architecture on Topic Maps is more natural and has advantages for their creation, usage and maintenance. In sum, subject-orientation and the Topic Maps paradigm have brought fascinating opportunities for portals. However, as always, much more work is ahead of us towards the further realization of the “seamless knowledge” vision for TOPs. The interested reader is invited to explore existing TMPs and the references provided.
KEY TERMS
Information Architecture: Information architecture is the application of knowledge organization principles to organizing Web sites by subject. In a Topic Maps-based portal, the Web site information architecture is driven by the knowledge model specified in the Topic Maps.
Knowledge Organization: Knowledge organization is the subject field concerned with ordering knowledge items (concepts) and the associated objects of all types relevant to these knowledge items.
Knowledge Portal: Aknowledge portal is an information portal used by knowledge workers to support them in their tasks. It is a specific type of enterprise portal, comprising support for information content storage and retrieval, organizational communication and group collaboration.
Ontology-Based Portal: An ontology-based portal is a portal employing ontologies as its semantic backbone, mainly for improved information integration, site navigation and search (querying and inferencing). In computer science, an ontology is understood as the explicit specification of a shared conceptualization (Gruber, 1993), or in other words, as a common model about things of interest in a particular domain people want to discourse about. In the case of a Topic Maps-based portal, the underlying Topic Maps specify a lightweight ontology.
Published Subject: A published subject is any subject for which there exists at least one published subject indicator. A published subject indicator is a subject indicator published and maintained at an advertised location for the purpose of supporting Topic Map interchange and merge-ability. A subject indicator is an information resource that is referred to from a Topic Map in an attempt to unambiguously identify the subject represented by a topic to a human being (Garshol & Moore, 2005; OASIS Published Subjects TC, 2003; Pepper, 2006).
Semantic Portal: A semantic portal is a portal using semantic Web technologies (such as RDF/OWL, or Topic Maps) to ease the (automatic) processing of content by computers.
Subject-Oriented Portal: see Topic-oriented Portal
Topic-Oriented Portal: A topic-oriented (or subject-oriented) portal (TOP) is a Web-based information or knowledge portal where informational content is presented in relationship to subjects. The information architecture is based on an ontology specifying the subjects of interest and their interconnecting relationships.
Topic Maps-Based Portal: A Topic Maps-based (or Topic Maps-driven) portal (TMP) is a topic-oriented portal which uses Topic Maps to realize the benefits of a subject-oriented information architecture.
