view printer-friendly version (opens in new window)
When metadata becomes content, and authoring learning
and the result an assessment method and part of a learner's/learning profile, you have something like Topic Maps. One of a number of emerging semantic web standards, it is the subject of quite a bit of research and development. The Norwegian estandard project and its members are busy working on educational applications of the technology.
Topic Maps are a means of capturing knowledge (as opposed to information) about a certain field. Rather than merely listing facts, it enables people to model how they think bits of information relate to each other.
The present ISO standard for Topic Maps was partially motivated as a means to merge book style indexes and seriously soup them up. That is still an important application of the technology, and not a bad way to conceptualise what they do: topics stand for meaningful concepts, and can point to occurences in a body of information, and be associated with each other, in a particular context.
Just like an entry for 'the internet' in a book's index could have a subentry for 'the web', so a Topic Map could be made where 'the web' is a topic, that points to, say, the World Wide Web Consortium (W3C) website, and is associated by an 'is instantiation of' link to the 'the internet' topic, all within a 'technology' context. In other contexts, you may well want to associate them differently.
Such a particularly crude example may not sound like much, until you realise that, unlike a paper index, there is no real limit to the number or depth with which such associations are made.
Better yet, each of the constructs in a topic map can be typed from an ontology that is itself a topic map: something like the 'is instantiation of' association above can be defined or qualified, and then used for inferences (depending on the tool).
Finally, because the things are expressed in a standardised XML format (amongst other formats), topic maps can be reused and exchanged. That holds the promise that a person's or community's understanding of a domain can be distributed, examined and processed much like flat information on the web today. Furthermore, given agreed subject identifiers, topic maps can be merged. Building a huge one collaboratively should therefore be possible.
Topic Maps as metadata
Fortunately, topic maps don't require formal semantics experts or internet revolutions to be useful. The job of giving a reasonably large load of learning objects some coherence and relevance to someone's course is already being done, and will remain a primary application.
In such use, Topic Maps are not likely to replace the IEEE Learning Object Metadata (LOM) metadata records that are the norm today. Though earlier, non-XML versions of Topic Maps did have a facility for recording simple property - value pairs (e.g. "interactivity: high") about the resources themselves, the XML version dropped these 'facets'. Mostly because it became much easier to point to actual online content rather than relatively empty subject indicators in the map itself. It is possible to dedicate topics to metadata about resources and associate them with a central topic that points to the resource, but there seems little advantage over a vanilla XML Schema LOM record.
In its use so far, Topic Maps seem more suitable for discovery functionality at a higher level: as a means to facilitate the structured discovery of resources for a whole course or sub-discipline, for example. Compared to the results list of the typical search, a topic map ought to provide a much more coherent, relevant and intelligible view of a topic, without the static and limited view that, say, a webapge full of links would provide. A readinglist with knobs on, in a way.
Topic Maps as content
Yet the use of topic maps in education goes beyond the functions usually associated with metadata. A hint of that is evident in the fact that the prototype TM4L ("Topic Maps for Education") editor has a facility for creating LOM records for the topic map itself. This is not surprising when considering that it is perfectly possible to make a topic map that essentially models a domain without any direct reference to resources outside of it.
As a content type, its use is akin to making notes when studying: a means to relate and process knowledge. The Brainbank application is mostly aimed at that kind of use by individual learners, and early indications of research into its use by secondary school pupils indicates that it does have a positive effect on such metacognitive skills.
The idea is quite persuasive: students can almost literally build up their knowledge of a subject by associating what they have learned from lectures or learning material in meaningful ways. An existing ontology can be provided, partially to guide, partially as scaffolding, and can even be challenged or adapted by the learner herself.
Unlike conventional paper notes, it's fairly easy to get back to parts you have done earlier and adjust them in the light of what you have learned later. Also unlike paper notes, it becomes easier to compare the understanding your peers have built up and borrow or merge any differences in a larger whole.
Topic Maps as assessment
This ability to compare topic maps also makes it suitable as an assessment method. A learner's understanding can be measured by comparing their topic map to a reference topic map of the same area.
In the Brainbank study, this was found to be effective, but quite time consuming. While that may partially be a user interface issue, it also follows a bit from the nature of the exercise. When comparing two mental models of the same domain, any comparisons are likely to require some thinking through. The conventional mental checklist model of assessing essays may be less thorough, but easier.
Topic Maps as a learning record
Building metacognitive skills, or "learning to learn" is a crucial aspect of personal development or lifelong learning records. An important aspect of that is the ability to relate new knowledge to what you know already. This includes prior knowledge across common subject divisions, and can involve very large amounts of information.
All of these things are part of using Topic Maps too: to build a topic map requires that new information be related to known information. If you keep at it, (and the tool can usefully handle it) the result is necessarily a constantly evolving accumulation of what has been learned.
Topic maps can structure large bodies of information in ways that are generally meaningful for a community, as well as particularly meaningful for the creator. They allow patterns to be replicated across domains.
One practical strength of semantic web technologies in general in the area of lifelong learning records is that they are flexible in structure, unlike the XML Schema based, fixed datamodels such as IMS Learner Information Profile. People do Personal Development Planning (PDP, the process that informs a learning record) in a myriad different ways, and mapping the stages and outcomes to a fixed datamodel has proven quite challenging.
A representational technology that can be augmented at will, and does not suppose that the tool that processes it knows in advance what the data will look like at every level could certainly help.
This quality of an instance carrying most of its own defining structure, is what topic maps have in common with other semantic web technologies such as the Resource Description Framework (RDF). This contrasts with the XML Schema based nature of most current e-learning specifications, which rely on central schema definition files (XSDs) for their structure.
The main difference between topic maps and RDF is that the latter is even more abstract and its predefined structure even more minimal. At the same time, its syntax is rather stricter. Also, where Topic Maps carry most of their own ontologies, RDF needs to rely on specific ontology languages such as RDF Schema or OWL.
The similarities and differences between RDF and topic maps is still hotly debated, but the bottom line is that RDF seems more generic, and therefore slightly harder to apply to a specific purpose. The greater generality also shows in the fact that, fortunately, topic maps can be expressed in RDF (given a good schema). The two technologies can be made to interoperate, as shown by the Ontopia Omnigator Topic Map browser that can display things like the RDF based Friend Of A Friend (FOAF) spec as a topic map.
Ifs and buts
The fact that there are such things as generic topic map browsers indicates an important, practical point about the present use of semantic web technology. The good thing about semantic web technologies is that they can capture knowledge in a flexible and powerful way, which is also the baed thing when it comes to designing applications that make use of them.
For XML and (X)HTML, the programme that you use to process data already knows the exact structure of the data that it will consume. IMS Content Packages, for example, will vary in some well defined ways, but a tool such as the Reload editor can know about and deal with all of these variations because it has access to a complete model of all content packages.
Semantic web technologies turn that model on its head. They are built to be able to capture an almost unlimited number of data structures, which means that applications that consume RDF or Topic Map instances can't really know what's coming at them beyond basic building blocks.
Topic maps are slightly better in this regard than RDF, because it is possible to know that much more about what each topic map will contain at a minimum. Even so, a generic topic map browser at the moment looks quite stark, and requires some skill and expertise to make sense.
With RDF, one pretty much has to constrain flexibility in advance in a particular application such as FOAF or RSS in order to be useable by a human looking at it in a software tool. Which is fine, but that approach loses the flexibility that is the key advantage over XML Schema.
For the use of topic maps in education, that means that, say, an assessment tool should present topic maps in a different way and with different controls from one that is meant for discovering resources in a repository. More than that, parts of a topic map that are used frequently, and play an important role in a particular domain (i.e. the domain ontology) ought to be presented in a specific way, without constraining the expressiveness of the topic maps unduly. And all of that preferably without requiring users to use squillions of tools that all look and feel completely different.
At the moment, it appears that there is still a lot to be done to solve that problem. Fortunately, there is a lot of work going on in the wider web world to enable tools that bridge the gap between the form & submit button website, and the traditional heavyweight desktop application that you need to install and learn. Somewhere in that world of browser toolbars and documents that come with their own user interfaces, there ought to be a solution that unlocks the power of topic maps for the average learner.
Most everything you wanted to know about topic maps in education is now being gathered in the Topic Maps in Education Wiki of the estandards project in Norway.