Towards a Semantic Infrastructure for Learning, Education and Training?
Position paper by Tore Hoel, Oslo University College, www.hoel.nu www.estandard.no
2007-11-11
Keyword: cetis-2007-conference-semantics
For Mac users the Leopard OS X brought boolean search and improved ability to map you own "time and subject flow" to the documents you have save through the years. Spotlighting for ' "learning technologies" AND (rdf OR "topic maps")' 216 documents turned up on my computer. In 2001 Mikael Nilsson told the readers of the CETIS web:
"The future for RDF for learning technology specifications is bright, and the possibilities opened up by RDF and Semantic Web technologies promise to take learning technology project to a new level of applications." (Mikael Nilsson, the CETIS web, September 27th, 2001)
The Cower Flow of documents on my computer shows that even if the future was bright, the progress in the learning, education and training domain for this technology has been slow. And we had to wait six years for this community to make semantic technologies an important track at the annual CETIS Conference.
In this short position paper I will share some of my thoughts on why we have experienced this slow uptake of semantic technologies. My context is of course Norway, the stronghold of Topic Maps, which is an alternative ISO technology to the ones coming out of the Darwinian race for semantic technologies (the W3C's RDF, RDFS and OWL, according the technology primer to this conference, ref. www.semantic-conference.com/primer.html). I will start with the big questions of how semantic technologies are aligned with our understanding of what Learning, Education and Training are actually about. I will then give some examples of Norwegian projects, i.e. how semantic technologies are used to defend the world view of "The Ministry of Truth", and how the very same technologies are used to help a local community to build the story of their cultural heritage.
The development of the semantic technologies are often explained by drawing trees or optimistic graphs pointing upwards - and only the sky is the limit for what we can do. (The drawing above is from a recent Norwegian comparative study of semantic technologies, focussing on RDF, Topic Maps, Core Components and ISO 15926 [1], (Aassve, 2007)). The 80's gave technologies for loose internal data coupling, the 90's for loose global document coupling (the web), the 00's gave SOA and loose process coupling, and now it is time for the semantic web with loose coupling of data and meaning.
We have seen lots of optimistic graphs promising a rich harvest of the higher hanging fruits beyond the taxonomies, the mind maps, the faceted classifications and the thesauri, as in this slide from the International Topic Maps conference in Oslo last year.
So why do these fruits not start banging on our heads? It might be that there are important results outside the radar screens of the ordinary university institution. Applications based on Topic Maps are used in Korea for classifying folk music; in New Zealand for navigation in an electronic text centre; in the US for National Nuclear Security Administration, and for keeping track of IRS taxation information, etc. And the swoogle.umbc.edu crawler finds more and more RDF documents on the web. However, in the LET domain we do not see much use of ontologies and technologies that allow us to harness these.
There are several reasons for this unsuccessful ability to pass the tipping point in our domain.
Firstly, there is a lack of tools and convincing demonstrators. It is a serious problem, but this is not the main reason why we hesitate to explore the benefit of these new technologies.
Secondly, as we move along the graph towards the applications that allow computers to make inferences based on formalised knowledge fragments, we realise that the institutions that rule our lives strongly resist to reveal and loose control of their knowledge structures. On top of Tim Berners-Lee's Semantic Web LayerCake there is the box of Trust. When we climb the EU layered model of interoperability and start to address other interoperabilities than the syntactical one, we are asking for trouble. Organisational (or political) interoperability is based on trust, which is hard to negotiate. And for semantic interoperability we need to negotiate meaning. Then you should better deal with institutions that do not see themselves as Ministries of Truth. Above all, building trust and open up the floor for negotiations take time, and it does not help to claim that you have slides showing how you will develop the software solutions.
Thirdly, the semantic technology leading the race talks nicely to machines, but does a rather mediocre job communicating the niceties of <resource, property, property-value> triplets across to the decision makers, or educators for that matter. While we are waiting for easy-to-use tools, we have seen that it does help to have a model that is easy to explain. It is my hypothesis is that the high penetration of this technology in Norway is due to the fact that it has been easy to explain the key concepts – the TAO of Topic Maps – to non technical people. Everybody understands that we have Topics, Associations between topics and Occurrences of these topics. Often that has been enough to sell a ontology driven portal i Norway.
Overall, the higher up in the tree you climb, the more irrelevant the technology is. It does not matter a big difference If you codify you ontology in Topic Maps or RDF. What matters is the effort to start describing your knowledge domain, and as the JISC Pedagogical Vocabularies Project (Currier et al., 2005) so clearly demonstrated, it is a long way to go just to start exchanging pedagogical vocabularies in the LET domain.
So may be we should start discussing what would be needed to establish a Semantic Infrastructure for Learning. Education and Training.
As long as I have been involved with learning technology standardisation I have observed an unlucky absence of pedagogical discourse. The best example is the unwillingness to discuss what is the scope of SCORM. We should realise that learning, education and training have different scopes. Training is often about getting a corporate (or other) view across. Education is more a rhetoric activity in a classical sense (Laurillard, 2002), often with a strong will of some Ministry of Truth to give an authoritative view of a domain. Learning, however, is another matter. The learners are often interesting in constructing their own knowledge, in cooperation with other fellow learners.
What does this mean for the development of semantic technologies? Ontologies are great for expressing fixed world views, e.g. curricula. But we do not believe in a transmissive model of learning, do we? So we need to develop semantic technologies that allow for different scopes or world views and facilitate negotiation of meaning. I am not saying that one flavour of semantic technologies is better than the other. I am only saying that these aspects (scoping and merging) are in the foreground of Topic Maps. And this is why Norway is such an interesting case to study, given the uptake of Topic Maps technologies in the educational sector in this country.
One of the Topic Maps success stories is the use of the technology for navigation in many governmental portals in Norway. The portal www.itu.no of The National Network for IT-Research and Competence in Education (ITU) was first. Topic Maps for navigation spread to the cultural sector (kulturnett.no), to the Government portal (regjeringen.no), a.o.; and lately to the Directorate for Education and Training. The Directorate is responsible for keeping updated versions of the curricula. These are pretty stable and structured documents that are well suited to be expressed as ontologies. The Directorate use Topic Maps to express the structure of these documents and allow for navigation in portals through re-use of their ontology through download in XTM format (XML for Topic Maps) or XML chunks. The updates are maintained using RSS feeds.
From our perspective it is interesting to observe how the locus of control is maintained in the Directorate's "GREP - living syllabi project". It is very clear that the syllabi are official documents with only one interpretation, - that of the Directorate. It has therefore been the primary concern of the Directorate to distribute the curriculum ontology "as is".
However, two mechanisms are designed to shift the locus of control to the users of technology. The first is a Norwegian convention initiated by the Norwegian eStandards project. In order to support organisational (and semantic) interoperability we proposed that every website should expose their "backstage", explaining what services they were running and how other implementers could make use of their APIs, RSS steams, etc. At bak.udir.no the Directorate explains how third parties could make use of the GREP ontology through downloads and RSS feeds. In this way the Directorate has open up a dialogue with the users how to build semantic technologies.
The other mechanism is part of the Topic Maps technology. Both RDF and Topic Maps identify a resource through a URI. The Topic Maps camp claims they have solved the "identity crisis" of the web [2] by distinguishing between using a URL as subject address to identify a information resource, and using a URL as a subject identifier, indicating what subject the URI is referring to. These fine distinctions we will allow to rest. However, what is important is the strong call this technology issues for establishing PSI servers, to resolve calls for Published Subject Indicators used as identifiers in Topic Maps applications. The PSIs are used to unleash one of the most powerful features of Topic Maps, the ability to automatically merge arbitrary topic maps.
If we study the internals of the Directorate's curriculum ontology we find identifiers both as URI references to PSIs and UUIDs. I find this illustrating of the the governmental ambiguity of keeping control vs. facilitating true dialogue. A UUID, e.g. 550e8400-e29b-41d4-a716-446655440000, is private and useless for discourse. A PSI, e.g. http://psi.mydomain.no/math , provided it resolves to a definition you may agree or disagree with, paves the road for exchange of meanings: We talk about the same thing or topic.
So, will the Directorate for Education and Training establish a PSI server for the curriculum vocabulary identifiers? It will, probably in 2008.
The next steps for a Norwegian infrastructure for semantic learning technologies will be to keep pushing for a more open professional discourse about the potentials of these technologies, stressing that every portal does have a "backstage" where the discussion on how to connect and make use of each others vocabularies, ontologies, APIs, etc. are continued.
From ten thousand feet ontologies of syllabi and other well structured domains seems a good idea. However, the problem is, as Simon Buckingham Shum puts it in his position paper, how to model contested domains. One answer to this challenge is to stop treating ontologies as final and closed structures that you need special tools to unlock.
The very reason why Topic Maps have become so popular for navigation in Norwegian portals is that the ontology is kept flexible and extendable during the development of the websites. However, the ontology itself is hidden from the users of these portals. And the navigation structures could as well be rendered from a relational database.
The small Norwegian company Cerpus has made two applications that put the ontology development right in the centre of the work process. Two years ago the CEO of Cerpus, Tommy Nordeng, and myself presented the "The Norwegian Topic Maps experience" to the CETIS Educational Content SiG i Manchester [3] demonstrating BrainBank Learning. BBL is still being used as a web application for learners to organise their concepts and resources constructing and co-constructing a topic map.
The new application is a community portal (kuling.net) for a local museum and regional community documenting the local history and coastal culture of Vesterålen in the Northern part of Norway. They are describing, mainly for the use in schools, local customs and legends in fishery, agriculture and local crafts and relates their texts, pictures and artefacts to places, time, landscape, seasons, etc. The portal is open for teachers and students to contribute texts and audiovisuals. However, the contributions have to be authorised by the museum staff.
When creating new text all users may extend the current ontology by adding new topics or associations. For instance, if the user is not satisfied with the subtypes of a geographical entity in the drop-down box, she may add a new topic to connect the text to.
The kuling.net application is still in beta. It is based on free and open software and should be a valuable contribution to the (slowly) growing number of semantic tools for learning. I do not think that this tool is exceptional in any way. However, the willingness to expose the internals both to the end users and to service implementers is pointing in the right direction. It is very likely that this method of knowledge and local heritage construction will spread to other regions of Norway and other domains, reusing some of the ontology that is co-constructed in kuling.net.
The uptake of semantic technologies in Learning, Education and Training have been slower than expected. This is not only due to the fact that these technologies are still in development; that the ideas behind the semantic web, RDF and OWL are complicated to deliver to the educational community, especially when we still lack good tools and demonstrators that give immediate benefit to learners and teachers. We have argued, however, that we need more than just better tools and technologies. If we think that the challenge is just to model authoritative knowledge domains, we will never reap the full potentials of these technologies. We should support knowledge constructions of ill-structured domains, focussing on the abilities of these technologies to support negotiation of meaning through comparing and merging learner constructed ontologies.
It is time for cross-national initiatives to explore the benefits of semantic technologies for Learning, Education and Training. To speed up the development we need to focus on both infrastructure (in a very broad sense) and applications. The following is my first points to a programme that this JISC CETIS conference could develop.
Aassve, Ø. et al. (2007) The SIM Report, A Comparative Study of Semantic Technologies
Currier,
S., L. M. Campbell and H. Beetham (2005) JISC
Pedagogical Vocabularies Project, Report 1
Pedagogical
Vocabularies Review
Laurillard, D. (2002). Rethinking university teaching - 2nd Edition. London, Routledge.
[1] ISO 15926 is a standard is to facilitate integration of data to support the life-cycle activities and processes of process plants. The data model defines the meaning of the life-cycle information in a single context supporting all the views that process engineers, equipment engineers, operators, maintenance engineers and other specialists may have of the plant.
[2] http://www.ontopia.net/topicmaps/materials/identitycrisis.html
[3] http://www.estandard.no/docs/ EC_SIG_Manchester_Hoel_Nordeng_2004-12-08.pdf