Web-based delivery of educational materials is a growing area in education, offering substantial support for teachers in the face of growing class sizes and declining resources. Additionally, such materials permit additional teaching aides, such as online discussion groups and bulletin boards, 24-hour availability, and in particular, a personalised lesson plan tailored to maximise a student’s benefit from the materials. This paper concentrates on the technical aspects of this latter feature, describing the implementation of adaptive hypermedia technologies in the WHURLE system and discussing the automatic generation of hypertext links in a body of educational materials.
It is well-documented that the online delivery of educational materials has many benefits (Laurillard, 1993). Primarily these are reducing the load of the teacher and being able to adapt the content of the materials, not just to the requirements of the lesson but to the individual strengths and weaknesses of the student. For example, some students learn more easily with pictures to supplement the text, while others are comfortable with symbolic representations. Online delivery also allows the materials to reach a far wider audience than a single classroom, resulting in a broader range of abilities and learning style, and potentially a diverse set of cultural backgrounds influencing the students’ comprehension, absorption and retention of materials.
Experimental evidence shows that the use of hypertext materials can polarise the benefit gained from online education (Quentin-Baxter, 1999) so that some students are systematically disadvantaged by the use of these materials while others perform much better than with non-digital materials. The aim is to overcome this disparity with an experimental Integrated Learning Environment - WHURLE (Web-based Hierarchical Universal Reactive Learning Environment) (Moore et al., 2001; Brailsford et al., 2002). It combines a number of technologies such as adaptive hypertext, computed linking, open hypertext and XML recommendations in order to create a learning system that tailors the organisation and presentation of learning materials to best suit the needs of individual students, using experimental user modelling (Zakaria et al., 2003).
The next section of this paper overviews the core relevant technologies that are a key part of a tailorable and interactive learning environment. The following section describes the implementation of such an environment, based on these core technologies, with the emphasis of this paper being on the linking technologies.
Related and Relevant Work
There are many areas of work which have contributed to the design and development of the integrated learning environment discussed in this paper and some of these are discussed in (Moore et al., 2001; Brailsford et al., 2002). This paper is focused on the hypertext linking as implemented in this learning environment, and this section will consider only those areas directly relevant to the linking work.
There are four major areas of work that have motivated and directed the work on hypertext linking reported in this paper. These are adaptive hypertext, which allows for the presentation of personalised materials to students, and computed linking, which automatically computes links for an arbitrarily-sized corpus. Other important technologies are the XML recommendations for interchange of arbitrary document formats, and open hypertext services.
2.1 Adaptive Hypertext
Fixed, or "static", hypertext provides the same experience for every person, regardless of their abilities, interests or browsing targets. The adaptive hypertext community aims to overcome the limitations of the “one size fits all” approach to delivery of reading materials by tailoring, or adapting, the content and the organisation of those materials to best fit the varying abilities, requirements and methods of the reader. It can also assist in ensuring that culturally appropriate materials are delivered to students from around the world (Stewart et al., 2003).
Adaptive hypertext is an especially useful technology for learning materials (Brusilovsky, 1998). Students begin learning from a diverse range of backgrounds, with inherently different learning styles and with different abilities. Adaptive hypertext allows different paths through the learning materials, tailored to the student’s needs. The student’s needs can be assessed as the student progresses, either by tests or analysis of navigation, or can be specified by the student themselves. This requires modelling the student and their abilities, to allow for tailoring of the materials and presentation according to need.
Adaptive hypertext systems generally use either the overlay model, which measures the learner's knowledge within a given domain (Carr & Goldstein, 1977; Eklund et al., 1997), or else the stereotype model which classifies students according to their background and abilities (Eklund et al., 1997). Zakaria et al 2003 have been developing a new, “hybrid” user model that combines these previous user modelling techniques, and is the one implemented in the interactive learning environment described later.
The reader is referred to (de Bra, et al., 1999) for a survey of adaptive hypertext technologies.
2.2 Automatic Generation of Hypertext Links – Computed linking
Automatic computation of hypertext links is accepted as a key tool for the generation and maintenance of hypertext materials, especially in a changing document corpus (Ashman, 2000) (Davis 1998). When the core materials change, the links connecting them become dislocated or irrelevant, or sometimes point to materials that no longer exist (Davis 1998). It is already a very time-consuming process to manually create a set of links for a body of materials, and to detect and correct any errors can add to this burden. A mechanism that automatically creates links and corrects any that may have been rendered wrong can save a hard-worked author or teacher a great deal of time (Ashman et al., 1997). This is especially so when there are a large number of links which could easily be specified with a simple computation but which are laborious to create manually (for example, links from names of historical characters to a biography, such as in (Fountain et al., 1992)).
However, in the educational context, another major benefit of automatic link creation is the tailorability that becomes possible by use of link computations that include some element of context. Adaptive hypermedia requires context to be taken into account when presenting materials and links to the student, and automatic computation of links using context variables as parameters to the computation is a highly efficient way to do this.
One side benefit of computed linking is that it can incorporate an arbitrarily large corpus of information from arbitrary sources. Links that are dynamically computed (as opposed to precomputed) can be created over any data from any source, merely by applying the computation to that data. This can be done in a “late binding” fashion (Brailsford, 1999) so that sets of appropriate links can be applied to the data immediately before it is made visible on the viewing tool. Dynamic computation can save on storage of individual links, which can be important when searching through large sets of links (Ashman et al., 1997).
The reader is referred to (Agosti et al., 1996), (Ashman et al., 1997) (Green, 1999) and (Wilkinson et al., 1999) for surveys and detailed explanations of computed linking technologies.
2.3 Open hypertext services
The hypertext community are increasingly favouring the provision of hypertext services by a hypertext engine that is external to the application that displays data. This is not to suggest that applications displaying data ought not have a hypertext capability but rather that any application can use supplementary hypertext features without the need for incorporating them into the application, or the need for storing hypertext links.
A core characteristic of open hypertext systems is that the links (or computation specifications for computed links) are not stored within the data being linked. This has a range of benefits, such as making it possible to link into and out of materials of arbitrary type and from arbitrary sources (Cawley et al., 1995), using computation capabilities from arbitrary applications (Verbyla et al., 1994), and the ability to use different links for different situations. These are key elements of a tailorable learning environment, making it possible to modify the ordering and presentation of materials.
The reader is referred to (Carr et al., 1999) for a brief survey on open hypertext systems and to (Brailsford, 1999) for a survey on links not stored in the data being linked.
2.4 XML and its recommendations
The eXtensible Markup Language (XML) is a collection of recommendations for the creation of markup languages for document interchange over the Internet. A key feature is that an author can create their own markup language to contain tags and actions specific to the requirements of the application under development. It can also allow developers to specify data structures to be used. XML is favoured for the development of Internet-based applications as it enables the exchange of data and actions specific to the application in a simple, HTML-like format. Most importantly, it is possible to create pointers into positions within Web pages without needing pre-existing NAME tags, and hence avoiding the need to overwrite a document in order to provide a link into a specific position within it. XML also allows the specification of “out of line” links which allow an author to create a set of links both into and out of documents that are not necessarily owned by the author.
XML contains three recommendations specific to linking, these being XLink, XPointer and XPath. The reader is referred to (De Rose, 1999) for a survey on the XML linking recommendations, and to www.w3.org/XML for up to date information about XML.
The WHURLE Framework
The previous section highlighted key technologies important to implementing the hypertext functions for an interactive learning environment that is tailorable to the needs of individual students. In this section, an implementation based on these technologies is described.
WHURLE is an experimental adaptive learning environment for the web. It uses XML to store and deliver educational content in atomic constructs called chunks. Chunks are the fundamental unit of information in the environment, and are the smallest possible fragment of conceptually self-contained information (i.e. where the component parts don't make sense in isolation). A chunk is usually small, such as a paragraph of text or an image, but it could be large - as in the case of an entire legal document. A captioned image would generally be a single chunk, because the caption is unlikely to make sense without the image.
A lesson is very simply a collection of chunks with hypertext links to provide pathways through the information. Teachers generate lessons by creating a default pathway through the available chunks, but this pathway can be modified by taking into account observations derived from the student’s user profile.
WHURLE automatically generates navigation to enable movement around the hierarchical structure of a lesson. Previously this was entirely done by the WHURLE software, but a recent development is to outsource some of this activity to a specialist, robust linking proxy system, GOATE (Martin et al., 2002).
The base unit of information is the chunk. Data is stored in chunks, although link computations can address words, phrases or other items within a chunk. A purpose-designed markup language, the WHURLE Chunk Markup Language (WCML) is used to structure chunks.
However, the end user does not ever see chunks as such, rather they will see the combination of chunks and links into a single seamless document or document series, such as a lesson. Teachers create lessons as default pathways through chunks by specifying a lesson plan, which is another XML file, this time specified in another purpose-built language, WLPML (WHURLE Lesson Plan Markup Language).
The lesson plan contains a hypermedia pathway through the entire collection of chunks. This is created by teachers using WHURLE. A lesson plan consists of a hierarchy of levels, each containing one or more pages. Pages consist of chunks transcluded by means of XInclude.
When the lesson uses adaptive technologies to adapt the content and presentation to the student’s needs, the lesson plan will be more complex as it must represent the conditions, or “dependencies” which determine the inclusion and ordering of chunks and links.
Autonavigation is essentially the use of forward, back, up, down and other buttons associated with travelling through the structure of a document. It is distinct from hypertext linking in that it depends on existing document structure and its purpose is to facilitate movement around the formal document structure, while links can move readers through arbitrary parts of the document or other documents. Autonavigation can be implemented using hypertext links (which makes it more difficult to distinguish between the two) but in WHURLE links and autonavigation are implemented in two quite different ways.
The teacher arranges pages into a hierarchy in a lesson plan, with a page being a combination of one or more chunks. Students can navigate through this hierarchy either breadth or depth-first, depending on their preference. However it can be time-consuming to create all the necessary links for forward, back, up, down etc, but also chunks can be included into different pages, so explicitly declaring these structure-navigating links may not be appropriate as the structure of lessons containing these chunks may vary.
To make the teacher’s job easier, and to allow the reuse of chunks in a position-independent manner, WHURLE uses an autonavigation system, which uses XSLT to generate the structural links that comprise the navigational components of the virtual document.
XSLT makes it possible to express the relationship between nodes or pages, both in terms of ancestry and sibling (end-to-end and side-to-side) relationships. By looking where the current page occurs within the lesson plan, WHURLE can generate autonavigation cues for previous, next, sibling and up nodes in a lesson plan.
WHURLE implements a robust system of node to node linking at the chunk level. All links are two-way, and they do not break (i.e. if a link target is removed then links to it are not rendered – moreover if the target is reinstated then its links will still be in place).
The links are stored in a linkbase. This is an XML file that is either created by a teacher and specified in the lesson plan, or it created by the student and is a part of their user profile. Individual links are represented by the element and the nodes (i.e. the link ends) are represented by the element. Nodes specify the part of the chunk that forms either the source or the destination of the link. Links can be either one to one (“single” links), one to many (“hub” links) or many to many (“plural”).
Note that these are externalised links, i.e. they are links which are stored separately from the data being linked (Davis, 1995). This is the same in principle to the XML “out of line” links, but is not a feature supported by HTML.
In order to avoid a substantial performance penalty on the server, the links are not incorporated into the WHURLE node tree, rather they are inserted by a proxy system called GOATE.
3.4 Proxy-based linking using Goate
GOATE is an experimental linking proxy whose primary purpose is to intercept all incoming documents before they reach the reader’s browser and to preprocess to document to insert appropriate links into the browser’s copy of the document.
Standard open hypertext systems promote the use of external links in order to avoid the difficulties inherent in personalising links over data owned by others (Davis, 1995). This is invariably done as a “late binding” of links computed or stored elsewhere with the target document (Brailsford, 1999). GOATE uses the same late binding principle to compute or retrieve, and then insert links into a document immediately before it is displayed on the browser.
Goate is a link server. There are two conceptual parts, one of them being a link consolidation service which gathers links from a number of sources, stores them in a common, low-level format (IDO) and handles rationalising different languages generally. The second part is about taking a collection of links for a given page and presenting them to the browser in a way that achieves the desired effect. It could be said the first part is acting as a link server for the second part. In theory though, other clients could access the link server part of a Goate proxy (and distributed Goate proxies could work on this principle).
In terms of link recomputation: Goate itself does not actually compute any link end-points. End-point calculation is done by Goate’s associated Language modules which interface with the system. The same IDOs are used until it is detected that they are no longer valid. This detection is based on the end-points changing (for embedded links the source changing is also requires the specification to be reconsidered) or the Language module declaring a change in its link base. When this happens, the old IDOs affected are discarded and new ones produced by Goate making a call to the appropriate Language module.
The second conceptual part of Goate is still performed every time a document is loaded. That is, even if the IDOs haven't changed since last time this document was displayed the process of working out how these collection of IDOs should be rendered is carried out in full.
Goate is currently specialised in rendering for the HTML environment. However, in terms of link source there is no particular bias to HTML or XLink. It is perfectly valid to embed links within HTML or XML syntax, although those links need not be accepted HTML links or XLink.
Non-embedded links are just as valid as embedded (and in fact are slightly easier to deal with). Language modules just declare that links exist, given a current 'context' of document displayed and the links that exist in the linkbase. The source and format of these links are invisible to Goate as the Language modules declare IDOs (or for this paper 'a common, low-level, link declaration format').
The GOATE proxy is described in detail in (Martin et al., 2002).
One of the new directions in the ongoing research and development in this project is that of localisation issues in the presentation and inclusion of materials in lessons. We noted in the introduction that there was not just a variety of student abilities and experience, but also potentially a diverse set of cultural backgrounds which could influencing a student’s comprehension, absorption and retention of materials. This requires more than just rendering materials in the appropriate language and script, and needs an awareness of the cultural mores and assumptions associated with that language. This is called “localisation” and is essentially to a sensitivity to the cultural context within which lessons are used. We are considering how educational guidelines vary from country to country, perhaps depending on the country’s religious beliefs, commercial position or political agenda. Also we intend to investigate whether a literal translation accurately conveys teaching materials, as substantial attenuation of meaning can occur when materials are translated, especially repeatedly (Stewart et al 2003).
This paper and ongoing work were performed within the EU Minerva project investigating Adaptivity and adaptability in online distance learning based on ICT.
Agosti, M. and Allan, J., (eds), (1997). Special Issue on Methods and Tools for the Automatic Construction of Hypermedia, Information Processing and Management, 33(2), Elsevier.
Ashman, H., Garrido, A. and Oinas-Kukkonen, H., (1997). Hand-made and Computed Links, Precomputed and Dynamic Links, Proceedings of Hypermedia - Information Retrieval - Multimedia ‘97 (HIM ‘97) conference, September 1997, pp 191-208.
Ashman, H.L., (2000). Electronic Document Addressing - Dealing with Change, Computing Surveys, 32 (3) September 2000, ACM.
Brailsford, D.F. (1999). Separable hyperstructure and delayed link binding, Computing Surveys, Symposium Issue on Hypertext and Hypermedia (eds H. Ashman and R. Simpson), ACM, 31 (4es).
Brailsford, T.J.; Stewart, C.D.; Zakaria, M.R. & Moore, A. (2002). Autonavigation, Links and Narrative in an Adaptive Web-Based Integrated Learning Environment, Proceedings of 11th International World Wide Web Conference, Honolulu, Hawaii, 7-11 May 2002. Available via: http://www2002.org/CDROM/alternate/738/.
Brusilovsky, P. (1998). Adaptive Educational Systems on the World Wide Web: A review of available technologies. Proceedings of the Workshop "WWW-Based Tutoring" at the 4th International Conference on Intelligent Tutoring Systems (ITS'98). Available via: http://manic.cs.umass.edu/~stern/webits/itsworkshop/brusilovsky.html.
Carr, B. & Goldstein, I. (1977). Overlays, a theory of modelling for computer aided instruction. A Technical Report. AI Memo 406, MIT, Cambridge, MA.
Carr, L., De Roure, D., Hall, W. and Hill, G. (1995). The Distributed Link Service, A Tool for Publishers, Authors and Readers, Proceedings of the 4th International WWW Conference, available at http://wwwcosm.ecs.soton.ac.uk/dls/link_service.html
Carr, L., Hall, W. and De Roure, D., (1999). The evolution of hypertext link services, Computing Surveys, Symposium Issue on Hypertext and Hypermedia (eds H. Ashman and R. Simpson), ACM, 31 (4es), 1999.
T. Cawley, H. Ashman, G. Chase, M. Dalwood, S. Davis and J. Verbyla, (1995). A Link Server For Integrating the Web with Third-Party Applications, Proceedings of the 1st Australasian World Wide Web Conference, Australia, http://www.scu.edu.au/aus-web95/papers/integrating/cawley/
Davis, H.C. (1995). To Embed or Not To Embed, Communications of the ACM, Vol 38 no. 8, pp 108-109.
Davis, H., (1998). Referential integrity of links in open hypermedia systems, Proceedings of ACM Hypertext '98, pp 207-216.
Davis, H., Hall, W., Heath, I., Hill, G. and Wilkins, R., (1992). Towards an Integrated Information Environment With Open Hypermedia Systems, Proceedings of the Second European Conference on Hypertext, ACM, pp 181-190.
De Bra, P., Brusilovsky, P. and Houben, G.-J., (1999). Adaptive hypermedia: from systems to framework, Computing Surveys, Symposium Issue on Hypertext and Hypermedia (eds H. Ashman and R. Simpson), ACM, 31 (4es).
De Rose, S. (1999). XML Linking, Computing Surveys, Symposium Issue on Hypertext and Hypermedia (eds H. Ashman and R. Simpson), ACM, 31 (4es).
Eklund, J., Brusilovsky,P. and Schwarz, E. (1997). Adaptive Textbooks on the World Wide Web. Proceedings of the Third Australian WWW conference, Lismore NSW, pp 5-9 July 1997. Available via: http://ausweb.scu.edu.au/proceedings/eklund/paper.html
Green, S. (1999). Lexical semantics and automatic hypertext construction, Computing Surveys, Symposium Issue on Hypertext and Hypermedia (eds H. Ashman and R. Simpson), ACM, 31 (4es).
Laurillard, D. (1993). Rethinking University Teaching: A framework for the effective use of educational technology. Routledge
Martin, D. and Ashman, H. (2002). GOATE: XLink and Beyond, Proceedings of Hypertext '02, ACM, 2002, pp 142-3.
Moore, A., Brailsford, T.J. & Stewart, C.D. (2001). Personally tailored teaching in WHURLE using conditional transclusion. The Twelfth ACM Conference on Hypertext and Hypermedia, August 14-18, 2001, Denmark. pp 163-164
Quentin-Baxter, M. (1999). Quantitative evidence for differences between learners making use of passive hypermedia learning environments, Computing Surveys, Symposium Issue on Hypertext and Hypermedia (eds H. Ashman and R. Simpson), ACM, 31 (4es).
Stewart, C., Brailsford, T. and Ashman, H (2003) Internationalization for Technology Based Learning Resources,23rd Internationalization and Unicode Conference, Prague, Czech Republic, March 2003.
Verbyla, J.L.M. and Ashman, H.L. (1994) A User-Configurable Hypermedia-based interface via the functional model of the link, Hypermedia 6(3), pp 193-208.
Wilkinson, R and Smeaton A. (1999) Automatic Link Generation, Computing Surveys, Symposium Issue on Hypertext and Hypermedia (eds H. Ashman and R. Simpson), ACM, 31 (4es).
Zakaria, M.R. and Brailsford, T.J. (2003) User Modelling and Adaptive Educational Hypermedia Frameworks, to appear, New Review of Hypermedia and Multimedia (Special Issue on Hypermedia and the Web), Taylor-Graham.