KWTR: Semantic browsing and learning
From semanticweb.org
[edit] Trends in theories and methods: short term (0-3 years)
As the search technology on the Web becomes more robust, the access to information in general, and to learning resources in particular, is improving. It is an old cliché that we are now facing information overload and new means are needed to reduce the amount of information we are interacting with. If we want to peruse the existing web technologies in the educational domain, we need to make a shift towards the “web is for exploration” paradigm. More specifically, we see two distinct modes of exploratory learning: (i) convergent, ‘spotlight’ browsing of internal resources, and (ii) divergent, ‘serendipitous’ browsing in an open web space.
The spotlight browsing approach to supporting exploratory learning is based on a cycle involving the selection of a collection of resources, the organization of this collection into pedagogically effective presentations, and the exploration of those presentations by the learner. It is suitable for the learning situation where the learner has a specific, welldefined objective. The concept in question becomes the centre of the spotlight, and the ‘shadow’ cast by the spotlight leads to semantically close or similar concepts.
On the other hand, the serendipitous approach to exploratory learning relies on the learner engaging in a different cycle: the presentation of an arbitrary, serendipitous resource, the discovery of domain-specific anchors in this resource, and the exploration of a range of other resources starting from the anchors. This model is more suitable for a learner who wants to have an overview of a domain or to learn something about the structure of a particular domain without any specific objectives.
Both models of exploratory learning lend themselves rather well to the emerging semantic web technologies. The common ground shared by both models is their flexibility, extensibility, and their potential for customization.
Moreover, the semantic web should be treated as an enabling and enriching technology, which can sit alongside and add value to the existing technologies (such as web browsers or digital libraries). However, in addition to developing rather passive extensions to the standard tools, an important aspect of the Semantic Web is the capability to put users/learners into a more active role, for example, as annotators of web content.
We also envisage that approaches to semantic browsing will go beyond the current single-ontology paradigm, to provide dynamic, opportunistic views of semantic content, to support browsing and exploration on the web. To achieve this goal several problems need to be addressed. For instance, it will be necessary to ensure that the user is not overwhelmed by the potentially huge amount of semantic information, which can in principle be relevant to the current web resource. It will also be necessary to develop reasoners able to assess the relevance of semantic information residing in distributed ontologies with respect to the current web resource and make decisions about which information to present and to what level of detail. These reasoners will be able to reason about the provenance of the information to assess its value with respect to the current browsing context and to perform mappings on the fly to integrate information derived from distributed, heterogeneous ontologies.
[edit] Trends in theories and methods: medium term (3-6 years)
Graphically oriented representations of domain knowledge such as argumentational networks, narrative structures or causal models will allow users to navigate both textual and graphical representations and provide access to complex models as well as alternative pages. As part of this work we envisage the development of a range of services associated with the building, maintenance and navigation of these complex representations. We also envisage a further range of community-oriented services and tools which will allow communities to express and transmit their community knowledge via these complex representations. We also expect that by this stage the semantic web will be a reality, with very large amounts of semantic markup available to support learning, interpretation and personalization. Hence, in addition to the issues of visualization and ontology mapping described earlier, issues of scalability and trust will also come to the forefront, as semantic information will be readily available and the tools will also be there to make use of it. On the (Semantic) Web, the large number of community Web sites and social networks make it difficult to choose and find the ones a community member needs to take part in. To assist community discovery algorithms, ontology matching techniques, and ways to aggregate and visualize information about communities need to be developed. Flink [Mika, 2005] is an example of current Semantic community portals addressing the challenge of aggregation, visualization and presentation of community information. Once the people, objects and processes are being annotated, and the Semantic Web is being easily extended by the communities of users and developers, delivery of massive volumes of Semantic content and workflows to the community members is a major challenge. The solution is expected to stem from the active research fields in the Semantic Web area. For example, Decker and Frank [Decker and Frank, 2004] (note 6) address this problem by combining the current Semantic Web developments in a Social Semantic Desktop, which will let individuals collaborate at a much finer-grained level than is currently possible, and save time on filtering out marginal information and discovering vital information. Delivery of community-driven Web content will also interoperate at a Semantic level with mobile devices, as first projects start to appear, e.g., Semapedia (note 7): an application of Web-based Wikipedia to mobile environments.
[edit] Trends in theories and methods: long term (6-12 years)
It is difficult to envisage what form the Semantic Web will have taken as we approach 2020. However, on the assumption that it will succeed, it is likely that it will remain composed of two essential features – ontologies and services – since these provide the means for capturing factual/relational and procedural knowledge respectively. It is likely that ASPL/Magpie [Domingue, Dzbor, Motta, 2003; Domingue, Dzbor, Motta, 2004; Dzbor, Motta, Domingue, 2005] will continue to expand its range of services and that these will become both more complex, user- or community-specific and supported by a plethora of novel computational and display technologies. We would hope that as these services become more and more usable (e.g., via voice activated, ubiquitous computational devices with holographic screens) they also become more firmly grounded in well-tested theoretical foundations. Thus, we would expect that services oriented to learning would combine a range of information retrieval technologies with the means for sifting, sorting and rearranging knowledge and expressing the results as part of some clear and clearly supported educational task while being fully and seamlessly integrated into the learner’s day to day activities.
[edit] Trends in tools: short term (0-3 years)
There are some ontology engineering tools available freely (such as Protégé) and commercially (such as Ontostudio). Another example is Oyster (which can be freely downloaded at http://oyster.ontoware.org) is a Java-based system that exploits semantic web techniques in order to provide an innovative and useful solution for exchanging and reusing ontologies, providing facilities for managing, searching and sharing ontology metadata in a P2P network, thereby implementing the OMV* (Ontology Metadata Vocabulary) proposal for the standard set of ontology metadata (http://ontoware.org/projects/omv/). In any case, most of them can still only handle a limited amount of data. Finally, the main tools in this area at the moment are social networking tools, such wikis and blogging. Most of these tools are free, even if some commercial ones are sold on the market at reasonable prices.
As we have already indicated, ASPL/Magpie operates as a learning support tool for learners wanting to familiarize themselves with the Semantic Web Studies domain. An earlier version of Magpie has already been successfully deployed as part of the Climateprediction project (http://www.climateprediction.net/) and is being used as part of the educational infrastructure for an Open University course on climate change. We have also had interest from online journal providers.
[edit] Trends in tools: medium term (3-6 years)
Extended semantic browsing tools (such as ASPL/Magpie) with a fully service based architecture with multiple ontology-supported services provide access to and navigation through graphical as well as textual resources. It is likely that the advanced platform for learning (ASPL) will be able to access any material available via the internet and to provide a personalized display of this material via fully tailorable web pages or other programmable display types. What is more, when we refocus the attention of the Semantic Web community on capturing the users’ statements rather than expecting users to do formal annotation, this opens a wide array of possibilities. One particular strand we are exploring at the OU concerns the role of mobile devices such as PDAs or telephones in learning. Since these devices are with the learners almost all the time, it is possible to start exploration/annotation e.g. through a mobile phone and later continue on desktops. Hence, a learner does not need to make a note of an interesting object (say a painting) to explore it at home. S/he can simply trigger the exploratory processes straight on the spot – in the gallery, in the café or in a train. This would obviously open further opportunities, e.g. embedding domain knowledge with geographic and positional knowledge, so that it becomes possible to customize the presentations of resources truly to the level of individual learners.
The People’s portal infrastructure [Zhdanova, 2004] allows end users to define the content structure (i.e., develop ontologies), populate ontologies and define the ways the content is managed on Semantic Web community portals where the People’s portal infrastructure is applied. Content management features on the People’s portal include ontology matching support, personalization support (at the personal and community levels) and dynamic reaching of a consensus on the basis of heterogeneous ontologies. The People’s portal was deployed as a part of an intranet at DERI – Digital Enterprise Research Institute [Zhdanova et al., 2005] and as an extension to the portal of a Semantic Web community (knowledgeweb on the people’s portal: http://people.semanticweb.org). The ontology matching part of the People’s portal was deployed as a Web application open to everybody on the Web (OWL Ontology Aligner: http://align.deri.org). In this respect, the People’s portal environment appeared to be planned from the very beginning to make a contribution to a trend that now proves to dominate in acquisition of the Web structures. Ontology acquisition from regular community members has not yet become a common practice on the Web, but current trends convince us that it will become among most common practices.
[edit] Trends in tools: long term (6-12 years)
ASPL descendants will be able to access, reason about, manipulate and (re)display any available web information. They will have immediate access to a range of ontology repositories as well as a range of mapping tools or services. ASPL itself will become a more pro-active environment in which communities continue to use ontology-oriented tools to express their knowledge. Indeed, since they are pro-active, ASPL and its descendants will become an integral part of these communities as they access, process, filter and suggest changes to ontologies, services, domain representations and content. ASPL or ASPL-like systems will become indispensable as a means of countering the massive amounts of information, misinformation and disinformation available via the Internet.
[edit] Trends in services and applications: short term (0-3 years)
[edit] Trends in services and applications: medium term (3-6 years)
Semantic browsing and learning portals will continue to be an invaluable means of providing computational assistance to learners, whether they are following a prescribed course or not, and to anyone battling the seemingly unstoppable flood of available information. For instance, they may be used by course developers as a means of accessing and arranging a set of resources into some narrative thread with central pathways and interesting but elusive byways. We are currently looking at how to abstract the structure of learning narratives in order to produce a high-level layer of semantics reusable within different domains. We are designing an application that allows the semantic annotation of philosophical resources, with the aim of supporting the automatic creation of learning narratives through the inserted material. As part of this framework, we are building a domain ontology covering fundamental philosophical concepts. The semantic relationships between these domain concepts will allow the formalization of specific learning narratives in a second ontology. So, for example, ways to browse this semantic space can be (at a high level):
- the critical explanation of a concept/theory (a learning path that highlights opposing theories, and the problems on which they are focused);
- the contextualization of a concept/theory (a learning path that shows associated information about an author, or the historical period, or other contemporary important theories in different research areas);
- the production of an author (a learning path that collects all the activities and results of an author, and organizes them according to user’s preferences);
- the intellectual lineage of a concept/theory (through a learning path that follows the influence of ideas across different areas and historical periods).
In a second phase, these results will be generalized and the framework extended to other subjects of educational courses, in order to define what the abstract features of a learning narrative are.
[edit] Trends in services and applications: long term (6-12 years)
Semantic browsing and learning systems descendants, in addition to their pedagogic role, will increasingly form part of all interactions with information. ASPL-like systems will initially be incorporated into ALL web browsers. However, as computing becomes ubiquitous and computational machinery becomes part of the everyday world with ondemand interfaces for input and display, ASPL-like systems will form an essential component of their infrastructure in providing contextualized, tailored information in a contextualized, tailored form. We will no longer interact with raw information – we will interact with the representations provided by ASPL-like services in a Gibsonian cyberworld which is at once pleasurable, productive and pedagogically sound.
