ATLAS: A Web-Based Software Architecture for Multimedia E-learning ...

ATLAS: A Web-Based Software Architecture for Multimedia E-learning Environments in Virtual Communities Marc Spaniol1 , Ralf Klamma1 , and Matthias Jarke2 1 2

RWTH Aachen, Informatik V, Ahornstr. 55, D-52056 Aachen, Germany {mspaniol|klamma}@cs.rwth-aachen.de Fraunhofer FIT, Schloss Birlinghoven, D-53754 Sankt Augustin, Germany [email protected]

Abstract. Multi-perspective problem solutions, leading to an increasing complexity in creating authentic learning scenarios and collaborative learning strategies, have gained the focus of scientific research. In order to assist learners in virtual communities working with digital media artifacts we analyze the needs of communities in different scientific domains ranging from the humanities to engineering. We combine our results with a media theory developed in Germany’s first interdisciplinary and collaborative research center on ”Media and Cultural Communication”. Based on the operational processes named transcription, localization, and addressing we introduce ATLAS, a web-based software architecture for multimedia e-learning environments in virtual communities. Further, we test metadata standards like MPEG-7 for digital media management in virtual communities. Exemplarily, we present the movie triage environment MECCA supporting an interdisciplinary community of scientists from the cinematic sciences, art history, and literature studies.

1

Introduction

For the past five years in our collaborative research center we have been studying the knowledge management strategies and learning processes in scientific communities in the humanities and engineering [KlJa99,BeKl01]. The center’s scientific community covers all sorts of scientists, from cinematic scientists to philologists. Our research has led us to the assumption that learning and collaboration among learners in the humanities has a more discursive nature when compared to engineering. This process of knowledge creation by interpreting and discussing selected phenomena is leading to scientific progress, but is hard to formalize with conventional computer science methods. Two aspects being recognized among these communities are: • The semantics of multimedia heavily depends on the discursive acts within the community of practice. • Communities within the humanities use multimedia artifacts frequently. W. Zhou et al. (Eds.): ICWL 2003, LNCS 2783, pp. 193–205, 2003. c Springer-Verlag Berlin Heidelberg 2003


194

M. Spaniol, R. Klamma, and M. Jarke

This phenomena consequently lead to challenges for computer scientists in managing and presenting information in networked information systems. First, an evolving terminology is needed based on discursive processes and differences in the usage of terms in an area of specialization. We are seeking for opportunities in organizing terms of conversations within communities like ontologies, which have been specially designed to structure and organize discursive content. For computer science this results in a challenge of managing views allowing duplicate, redundant, or even contradicting descriptions to co-exist. The second aspect is the need for using multimedia artifacts for information presentation. Due to the ability of computer science to (re-) combine and organize digital media, we support scientists from the humanities with various fields of specialization. Learning is a social system within the communities of practice [Weng98] that needs a tight interplay of communicative acts and the organization of knowledge. We have realized the need for computer supported communities in creating their evolving terminology of conversations in discursive processes. In a distributed setting this leads us to virtual communities and ontology management systems for virtual communities of learners. To satisfy communities’ needs we have to find an opportunity in combing ontology management systems with multimedia artifacts as carriers of learning content. Consequently, virtual communities of learners need support in digital media processing that allow them to add multimedia with high-level semantics. In this paper we’d like to present and discuss different approaches for multimedia information processing in e-learning environments. The rest of the paper is organized as follows: In the next section we introduce the emerging theory of the community of practice we are cooperating with in our collaborative research center consisting of the concepts: transcription, localization, and (re-) addressing. Then we analyze current strategies on multimedia management in e-learning environments. Afterwards, we present the MECCA case study, a multimedia e-learning environment based on MPEG-7. The paper closes with a summary and an outlook on further research.

2

MM Management Strategies in E-learning Environments

The diversity of media allows learners to select information from various sources as well as to create content in heterogeneous formats. Especially, new media allows an even faster and more complex structuring (re-)configuration of information, making it more difficult for users to find the best fitting data set. When developing e-learning environments we are focusing on optimizing addressing of information in terms of content adaptability and functionality of the user interface. In contrast to approaches using low-level features by extracting semantic content as color, sound, or shape [Ciep01,SpFa02], humanist scientists communities need high-level semantic annotations to manage multimedia. There is also some research done on semantic indexing of multimedia [GrSr01], but it is missing features to manage divergent terminologies as it is needed for multimedia views. To overcome these challenges our colleagues from the humanities

ATLAS: A Web-Based Software Architecture

195

Fig. 1. Domain specific processing of MM files in ontology-based information systems

have developed the theory of transcription, localization, and (re-) addressing of multimedia. We are giving an overview on how these well-defined terms can be interpreted for the usage in computer science [Jark02]: • Transcription is an operation to make media settings more readable [JaSt02]. • Localization is a transfer of global media into local practices [Fohr03]. • The term of (Re-) Addressing describes an operation that stabilizes and optimizes the accessibility of global communication. Using these operational modes we will now analyze multimedia management strategies in e-learning environments. Basically, there are two strategies. Text and multimedia databases are approaches to handle data in a more or less unstructured manner in contrast to strict ontology management techniques. 2.1

Ontology-Based Information Systems

Ontology-based information systems have been developed to structure content and support information retrieval. They reach from simple catalogs to information system ontologies using full first order, higher order or modal logic [SmWe01]. Ontologies are based on modelling and abstraction of real world features [SSS*02]. The aim is to find a core ontology that can be modified to comply with specific settings [Guar98], which often results in a lack of flexibility since the structure of an ontology is too strict. For that reason, these systems are unpopular in some fields of applications, particularily where a somewhat individual classification of data is preferred. Experiences with other multimedia ontologies, like Dublin Core, lead to the insight that the effort of harmonizing relatively small ontologies often appears frightening and generates questions about scalability [DHLa03]. Surely, there is no alternative to merging of ontologies, which raises new scientific challenges. It requires considerable intellectual effort and is a learning process for individuals as well as their communities. The problem is that an ontology has to fit into all user interpretations, which becomes obvious when an ontology creation is shared [DCGR98]. Hence, developing an ontology is usually guided by domain experts in an iterative, incremental and evaluative process. This is commonly done by an ontology engineer designing a common ontology by assessing users’ and their community’s needs. The

196

M. Spaniol, R. Klamma, and M. Jarke

Fig. 2. Domain specific processing of multimedia information

understandability of the resulting ontology is questionable, since the used terminology is often (mis-) leading due to the user’s field of specialty. Commonly, the lowest common denominator of terms has to be chosen. Overall, ontologymanagement systems mainly support the exchangeability of terminologies to find a common level of conversation, but lack an integration of multimedia content. For that reason, information-brokering systems have been built on top of them. They combine the advantages of strict ontologies and support multimedia retrieval. Still, the remaining problem is that information-brokering systems lack flexibility in a multi-user and distributed ontology creation process. By means of figure 1, creation, organization, and presentation of e-learning content in an ideal ontology-based information system is being explained. Transcription (1) is used to make multimedia content better understandable for others. This can be done by annotating MM files and storing them e.g. in XML. The semantic enriched data are now ready for further processing. Now, a domain specific ontology is being used to structure the needed data. Therefore, an ontology engineer creates a localized (2) ontology assuming the needs of users in a specific domain. Semantic enrichment and categorization of MM files is performed in a second transcription (3) process, since the data has to be interpreted domain specific. All files are now localized (4) in an ontology management system (OMS) where they are stored for further processing. Depending on the user’s interests - stored in an ontology related scheme - the MM files are presented to the user in a final (re-) addressing (5) process.

2.2

Ontology-Based Multimedia Management by MPEG-7

The MPEG-7 metadata standard [Mart02] is XML based and has been introduced by the moving pictures expert group (MPEG) [BYFe02]. MPEG-7 has

ATLAS: A Web-Based Software Architecture

197

Fig. 3. Processing of MPEG-7 multimedia data by an underlying MPEG-7 ontology

been designed to describe multimedia content of different datatypes. Basic concepts of MPEG-7 are descriptors and description schemes (DS). Descriptors define syntax and semantics of each feature or metadata element, whereas the DS specifies the structure and semantics of the relationships between components. Additionally, the description definition language (DDL) allows the creation of MPEG-7 descriptors and DS. It provides a syntax to combine, express, extend and refine descriptors and DS [Hunt01]. MPEG-7 offers few means to manage content by strict formalizations just as ontologies do. Nevertheless, content management and personalization in an ontology-like structure can be managed by MPEG-7 (at least) twofold. The box on the left of figure 2 covers elements inherent in MPEG-7. They include the DDL, DS and descriptors, which can be freely rearranged. Pursueing a graphbased multimedia management strategy can be done fully compliant to the descriptors inherent in MPEG-7 (”MPEG-7 box” in fig. 2). Another possibility is to use structural aspects of the description schemes by interpreting the underlying tree-hierarchy and their nesting as a domain specific ontology. Usually, this requires the combination of newly defined DS and descriptors with those predefined by MPEG-7. Figure 2 indicates the latter approach as domain specific extensions outside the box capturing MPEG-7 inherent elements. Since these elements are also defined by the DDL and handled like those elements that are MPEG-7 inherent, validation and consistency checks can also be performed with those newly created components of a MPEG-7 schema. Since we are trying to achieve the best possible exchangeability of multimedia content we are developing MPEG-7 environments fully compatible to the basic set of MPEG-7 descriptors. This makes an exchange of content possible without further (mis-) interpretation of additional defined DS and descriptors. Figure 3 clarifies the processing of multimedia content by an ontology stored as MPEG-7 graph. Based on a global view on multimedia (MM) files to be managed a localized (1) ontology is being developed. A core ontology is defined a-priori or by a collaborative ontology creation process [KSJa03] by interpreting (transcribing (2)) the MPEG-7 graph as an ontology. For a better understanding of the content MM files are transcribed (3) within the MPEG-7 descriptors. That allows us to combine the advantages of structural organization of content with unclassified metadata annotations. After that, the data is localized (4) in an

198

M. Spaniol, R. Klamma, and M. Jarke

Fig. 4. MPEG-7 graph representation

Fig. 5. Ontology stored in a MPEG-7 graph represented as a tree

XML database. From there it is accessible to all users in a final (re-) addressing (5) process using the classifications within the MPEG-7 graph as well as free text metadata annotations.

3

Building and Maintaining Ontology-Based Multimedia Management Systems

Adopting global media to a community’s needs is the crucial aspect of localization, i.e. making media accessible and understandable for community members in their current practice. As we have discussed in the previous section, an ontology covering the terminology of the domain is needed to manage multimedia content in both cases. MPEG-7 is advantageous with respect to localization, transcription, and (re-) addressing, since it is capable of supporting all of it. At the top of figure 4 a DS responsible for localization in MPEG-7 ontologies enables us to integrate heterogeneous media. In general, we are able to integrate global media to local settings. What comes next are transcription features. They are used for adopting global media for a better understandability. Here, we are making use of standard MPEG-7 DS and descriptors that carry additional information in form of metadata in predefined tags. Finally, addressing in MPEG-7 is supported by

ATLAS: A Web-Based Software Architecture

199

Fig. 6. Personalization and context adaptation in MPEG-7

an ontology being represented by the DS and descriptors in the lower section of figure 4. Exemplarily, we have filled in some nodes and relations, which are framed by the superordinate MPEG-7 graph description scheme. The core of our MPEG-7 ontology multimedia management system is a MPEG-7 graph representation. In addition to media related metadata it consists of three components (cf. figure 4): • The graph description scheme • The descriptor(s) for node elements • The relation description scheme The graph description scheme contains all relevant information to define the structure and the elements of an MPEG-7 graph. It serves as an overall frame tagging all its sub elements. In general, nodes in MPEG-7 are optional. Any item referenced by a relation in a graph is automatically defined as a node (this is called an anonymous node). For the sake of clarity we have chosen the opportunity to define nodes directly. The descriptor for node elements defines the structure of the corresponding node elements. Each element is defined by its ID attribute. The relation description scheme can be used in two ways. One format is an external one that contains the relation apart from their related descriptions of the description scheme instances. The other format is an internal one that embeds the relation directly within the description of the source argument of the relation. In this case we have chosen the second opportunity. Figure 4 shows the application of the previously described technique with project specific annotations of the multimedia artifacts. Our internal relation description scheme specifies, within its relation elements, the type of the relation as well its source and its target attributes indicating the nodes joined by the relation.

200

M. Spaniol, R. Klamma, and M. Jarke

The ontology represented as graph supports users navigating through digital media. In our case, we are using tree structures as s subset of a graph. Figure 5 shows a sample tree structure of an application ontology as a result of a discursive ontology creation process among cinematic scientists. The partially spanned tree of figure 5 indicates, by grey color, the corresponding categories of multimedia file. A related MPEG-7 document describing multimedia content (user has been made irrecognizable for the sake of privacy) is being shown in figure 6. All selected categories are represented in the body of the document. In order to support various communities, we are introducing an architecture supporting high-level semantic annotations of multimedia artifacts based on MPEG-7.

4

ATLAS: A Web-Based Community Software Architecture

ATLAS (Architecture for Transcription, Localization, and Transcription Engineering System) is a community management system handling multimedia artifacts in networked communities. The features of ATLAS are shown in figure 7. Based on our theory as a modification of Nonaka and Takeuchi [SKJa02], learning takes place when we successfully internalize the transcribed knowledge that has been created within our community. Creating new content for the community is done by writing reports, homework essays, etc. These multimedia artifacts are managed in a community repository together with community relevant information while all the metadata are stored in an MPEG-7 compliant XML repository. ATLAS components use both repositories. The measuring component is constantly assessing community needs by gathering and evaluating quantitative data. Searching, Browsing, and Personalization are components of the transcript engine, which allow a community to comment on existing media by using other digital media and thus addressing their own needs or stabilizing media addresses. This is done by metadata supported semantic zapping. Semantic zapping means metadata-mediated browsing, allowing easy access to semantic information by supporting retrieval of multimedia artifacts according to the learning task. Strategies for efficient content management are crucial for e-learning environments since the amount of data can rapidly reach critical sizes. On the one hand there is the need of extensive and attractive multimedia presentations, but on the other hand network traffic should be reduced to speed up queries. Hence, we are applying a mix in storing the digital content on the client-side as well as on the server-side. On the client-side we keep the multimedia content, e.g. movies, on hard disk or on CD’s that might cause disorders when accessed via the world wide web. On the server-side we have set up an Apache webserver situating an XML-database, which contains metadata on the digital content and associated collections compatible with MPEG-7. For that reason, we can easily transfer our content to and from other MPEG-7 based applications. User requests, e.g. by XQuery, result in an XML-file sent to the client. It contains the concerning metadata of the digital content to be presented in a Java-based MPEG-7 player. Since the implementation has been done in Java, ATLAS is platform independent

ATLAS: A Web-Based Software Architecture

201

Fig. 7. ATLAS community software architecture

in general. The overall architecture is an extension of Grosky’s metadata mediated browsing [GrSr01] and research on more detailed and performance oriented eventually layered digital media management architectures as in [BBH*02].

5

Experiences in Ontology-Based Multimedia Management

In our collaborative research center we are investigating the impact of digital media on learning processes in the cultural science communities. We are developing multimedia environments to make the learning process successful for both individuals and the community. Due to the interests of cinematic scientists in transcribing and commenting on multimedia artifacts, we are trying to combine those practices with methods of computer sciences as e.g. abstraction and categorization. Because of this, we have started a cooperation with cinematic scientists by jointly developing new information systems based on MPEG-7. This means that we have to reduce (or even close) the gap between unclassified semantic enrichment of multimedia and strict categorization of ontologies. The MoviE Classification and Categorization Application (MECCA) is a high-level semantic annotation tool for multimedia artifacts. It serves as a multimedia environment for online video triages and collaborative ontology creation, which is a discursive and multistage process. The systems is based on MPEG-7 in order to enable scientists to develop concurrent classifications based on a distributed setting and to keep flexibility in multimedia content to be integrated. MECCA is being used in the cinematic sciences covering users having diverse educational backgrounds, like cinematic science, history of art, or graphical design. This community brings together users having various levels of profession such as full professors, research assistants, or students. Community members

202

M. Spaniol, R. Klamma, and M. Jarke

Fig. 8. E-learning triage environment for cinematic science students (German version)

have different interests and point of views due to their educational background. In MECCA, users first take triages on the already existing multimedia content. In addition, users can add content compatible with MPEG-7. The next step is done by gradually annotating and classifying the data. Each users’ classification scheme is kept in a separate MPEG-7 file. To retain the semantics of a multimedia file individual annotations and classifications are possible. This means that we allow redundant, overlapping, or even divergent views. These personal collections can be distributed and discussed among other community members. To detect differences between individually created ontologies and those of their community the system now checks the structures represented in the MPEG-7 graph representation based on tree-comparison algorithms. Concepts matching fully or partially can be detected as well as those showing divergence. In addition, MECCA allows users to reflect the decisions that have been made by back tracing as in [CTZa02]. MECCA is based on the constructivist learning environment “Berliner sehen” [Fend01]. Learners are stimulated to freely explore the content led by high-level semantics of the MPEG-7 content description and share their experiences in collections with others via the web. The front end of our video triage application (cf. figure 8) is an enhancement of its predecessor, the Virtual Entrepreneurship Lab (VEL) [KHJ*02]. Main improvements are an increased flexibility of the cat-

ATLAS: A Web-Based Software Architecture

203

egorization panel and the integration of a search engine. The adaptive navigation toolbar allows displaying the underlying structure with respect to the cardinality of categories and components in the classification schema. To serve different classification systems further presentation styles from one-dimensional buttons to trees as well as combination of both can be selected. The search engine allows a comprehensive search on the users’ collections or on the original content.

6

Discussion

We have received positive feedback from the scientists who used our e-learning environments because it allowed them to comment media on media. Based on that feedback, ATLAS has been designed to comply with this task. Due to discursive knowledge creation processes in communities of humanist scientists, there is an interest in exploring distributed classification processes. About one year ago, the MECCA project has been introduced to our colleagues from the humanities. Starting with meetings of 6-8 members, the community initially defined a classification scheme on a drawing table. Members liked to define a common vocabulary but some terms have been critically discussed since community members’ disciplines cover a wide range. Hence, the overall community has rejected some terms since their interpretation might have been misleading to them. On the other hand, some special terms have been taken into the common classification scheme to allow specialists to classify the content in detail. These terms are not conflicting with the intuitive understanding of others, but due to their degree of specification into a subsection of the cinematic sciences, these patterns are rarely used. Hence, researchers are hoping that a computer-mediated system could detect those conflicts more accurately. Another aspect is that the hierarchy of professors, researchers, and students has been subliminal affecting the negotiation process. Hence, those members of the community being at the bottom of the hierarchy or belonging to a minority within the community are hoping that a computer-mediated classification process might give them a greater chance to push their interests. Similar environments for communities in other areas of application - like plastics engineering - are currently being developed on the basis of ATLAS. All our environments allow the transcription of content in a guided but not prescripted way. By using ATLAS, a speedier software transfer in different areas of applications on the base of MPEG-7 has been made possible. The use of MPEG-7 has been crucial for the development of ATLAS. It allows users to express high-level multimedia semantics in a collaborative knowledge creation process. In addition, heterogeneous information can be contextualized since MPEG-7 gives us the opportunity to manage content of arbitrary digital media formats.

7

Conclusions and Outlook

We presented and discussed current approaches on digital media management. In our case studies we demonstrated that the use of MPEG-7, as a common ontol-

204

M. Spaniol, R. Klamma, and M. Jarke

ogy language both for the multimedia artifacts and the community vocabulary, allow a more authentic, transparent, and flexible knowledge creation process. We implemented a software architecture called ATLAS supporting application building by common services and unified repository handling. In making such environments compatible to the MPEG-7 standard, the maturity of digital media management increases. Yet, hosting virtual communities of learners is still in its infancy. For best exploitation of explicit multimedia semantics, we are currently researching on accessorily options that MPEG-7 offers. Next steps include artifact transformation with parameterized XSLT (extensible stylesheet language transformation) scripts in our system. Another field of research is the implementation of an MPEG-7 hyperlink structure allowing to express and comment on relations of media files. Further research aims also at integrating streaming technologies into our learning environments. For that reason, we are currently investigating an integration of our Java applications in common browser technologies. Acknowledgements. This work was supported by German National Science Foundation (DFG) within the collaborative research centers SFB/FK 427 “Media and cultural communication” and SFB 476 “IMPROVE”. We’d like to thank our colleagues for the inspiring discussions.

References [BBH*02] Berthold, H., Binkowski, F., Henrich, A., Hollfelder, S., Lindner, W., Marder, U., et al.: Architektur multimedialer Informationssysteme. Informatik Forschung & Entwicklung 17(2) (2002), pp. 77-89 (in German). [BeKl01] Becks, A., Klamma, R.: Kooperative Dokumentanalyse in einem interdisziplin¨ aren Forschungskolleg. Schnurr, H.-P., Staab, S., Studer, R., Stumme, G., Sure, Y. (eds.): Beitr¨ age der Konferenz ’Professionelles Wissensmanagement – Erfahrungen und Visionen’, Baden-Baden, 14.–16. March (2001), pp. 289–307 (in German). [BYFe02] van Beek, P., Yoon, K., Ferman, A. M.: User Interaction. Manjunath, B. S., Salembier, P., Sikora, T. (eds.): Introduction to MPEG-7 – Multimedia Content Description Interface, John Wiley & Sons Ltd. (2002), pp. 163–175. [Ciep01] Ciewplinski, L.: MPEG-7 Color Descriptors and Their Applications. Skarbek, W. (ed.): Computer Analysis of Images and Patterns, 9th International Conference, CAIP 2001 Warsaw, Poland, September 5-7 (2000), pp. 11–20. [CTZa02] Chien, S.-Y., Tsotras, V., Zaniolo, C.: XML Document Versioning. SIGMOD Record Vol. 30/3 (2001), pp. 46–53. [DCGR98] Dieng, R., Corby, O., Giboin, A., Ribi`ere, M.: Methods and Tools for Corporate Knowledge Management. Proceedings of the 11th Banff Workshop on Knowledge Acquisition, Modelling and Management, KAW’98, Banff, Alberta, Canada (1998). [DHLa03] Doerr, M., Hunter, J., Lagoze, C.: Towards a Core Ontology for Information Integration. Journal of Digital Information, Volume 4, Issue 1, (2003). [Fend01] Fendt, K.: Contextualizing content. In Knecht, M., v. Hammerstein, K. (eds.): Languages across the curriculum. National East Asian Language Ctr., Columbus, Oh (2001), pp. 201–223.

ATLAS: A Web-Based Software Architecture

205

[Fohr03] Fohrmann, J.: Die Verfahren der Medien: Transkribieren – Adressieren – Lokalisieren. In: Fohrmann, J., Sch¨ uttpelz, E. (eds.): Die Kommunikation der Medien, T¨ ubingen, Niemeyer (2003), (in German). [GrSr01] Grosky, W.I., Sreenath, D.V.: Metadata Mediated Browsing and Retrieval in Semantically-Rich Cultural Image Collections. Proceedings of the 2001 Tokyo Symposium for Digital Silk Roads, Tokyo, Japan, December 2001 (2001). [Guar98] Guarino, N.: Formal Ontology and Information Systems. Proceedings of FOIS’98, Trento, Italy, 6–8 June 1998 (1998), pp. 3–10. [Hunt01] Jane Hunter: An Overview of the MPEG-7 Description Definition Language (DDL). IEEE Transactions on Circuits and Systems for Video Technology, 11(6) (2001), pp. 765–772. [Jark02] Jarke, M.: Wissenskontexte. K¨ unstl. Intelligenz, Heft 1/02 (2002), pp. 12–18. [JaSt02] J¨ ager, L., Stanitzek, G.: Transkribieren - Medien/Lekt¨ ure. Wilhelm Fink Verlag, Munich (2002), (in German). [KHJ*02] Klamma, R., Hollender, E., Jarke, M., Moog, P., Wulf, V.: Vigils in a Wilderness of Knowledge: Metadata in Learning Environments. Proc. of the IEEE Intl. Conf. on Adv. Learning Technologies, ICALT 2002, Kazan, Russia (2002), pp. 519– 524. [KlJa99] Klamma, R., Jarke, M.: A Comparison of Engineering and Cultural Science projects. ESCW-Workshop XMWS’99, Beyond Knowledge Management: Managing Expertise, Kopenhagen, Denmark (1999). [KSJa03] Klamma, R., Spaniol, M., Jarke, M.: Collaborative capturing and use of multimedia semantics in MPEG-7 based knowledge management systems. Submitted to: IEEE Multimedia (2003). [Mart02] Martinez, J. M.: MPEG-7 Overview. Intl. Organization for Standardization, http://mpeg.telecomitalialab.com/standards/mpeg-7/mpeg-7.htm, (2002). [SKJa02] Spaniol, M., Klamma, R., Jarke, M.: Semantic processing of multimedia data by MPEG-7 for capacious knowledge management. Grosky, B. (ed.): Proc. of SOFSEM 2002 Workshop on Multimedia Semantics, Milovy, Czech Rep. (2002), pp. 56– 65. [SmWe01] Smith, B., Welty, C.: Ontology: Towards a new synthesis. Welty, C., Smith, B. (eds.): Formal Ontology in Information Systems, Ongunquit, Maine; ACM Press (2001), pp. iii-x. [SpFa02] Spevak, C., Favreau, E.: Soundspotter – A prototype for content-based audio retrieval. Proc. of the 5th Intl. Conf. on Digital Audio Effects (DAFx-02), Hamburg, Germany (2002). [SSS*02] Schmitz, C., Staab, S., Studer, R., Stumme, G., Tane, J.: Accessing Distributed Learning Repositories through a Courseware Watchdog. Proceedings of the E-Learn 2002 – World Conference on E-Learning in Corporate, Government, Healthcare & Higher Education (2002). [Weng98] Wenger, E.: Communities of Practice – learning, Meaning, and Identity. Cambridge University Press, Cambridge, UK (1998).