The economic, pedagogical and technical needs of taking into account these various
aspects and the relevance of using semantic grids to do that has been recognized
by various researchers, e.g., Allison & al. (2005) and Gouardères & al. (2005), many
of which participate to ELeGI ("European Learning
Grid Infrastructure"), an European project which began in October 2004.
The semantic grids combine:
- Grid Computing techniques (these techniques allow the shared exploitation of the machines on the grid network, they offer a common software architecture and thus support the development of common services and, finally, like peer-to-peer networks they facilitate various services for data sharing services and collaboration);
- Semantic Web techniques - or more generally, techniques for representing and sharing knowledge - which facilitate the retrieval and combination of data and services.
However, the research domain of Semantic Grids is new (De Roure & al., 2003), and the Grid Computing techniques as well as the knowledge representation and sharing techniques are themselves far from being mature. Research in this domain is thus varied; here are some examples: 1) the exploitation by Allison & al. (2005) of the OGSA ("Open Grid Services Structures") platform to extend and apply the specifications of standards such as IMS-LD for the re-use and interoperability of learning objects, and hence, more generally, to specify and take into account the context of each student (e.g., identity, preferences, levels in various fields, characteristics of the interfaces and tools of communication own by each student), 2) the creation of the S-OGSA ("Semantic-OGSA") plateform by the OntoGrid project (an eight-partner European project launched in October 2004) to facilitate knowledge representation and sharing, 3) the CoAKTinG project ("Collaborative Advanced Knowledge Technologies in the Grid") which facilitates collaboration and data exchange during or after virtual meetings on a grid), and 4) GRID-e-CARD (Gouardères & al. 2005), a project which manages a model of qualification (or certification) for each learning object and for each student on a grid, in order to facilitate the learning of each student (in particular via her insertion in relevant communities).
However, in all the researches on semantic grids, and also in those on
Learning Objects (LOs), knowledge representation and sharing is actually much more
superficial than a surface reading of articles about those researches could lead us
to believe. For example, currently existing LOs are almost never about only one
relation (i.e., a current LO does not describe a relation between one object and
another, as for example a relation between the "cosine" function and a definition for it
or a theorem using that function, or a relation between the "Java" language and one
of its characteristics) but to a set of objects (for example, a typical LO about Java
is an "introduction to Java" which describes the main characteristics of Java and gives
an example of code). This goes against the theoretical goal of LOs which is to
obtain modules of information as small as possible in order to increase the choice and
possibilities of re-using and combining these LOs. Moreover, the metadata that can
currently be associated with these objects are poor and informal (thus, hard to exploit
automatically); typically: the name of the LO's author, an informal description of the LO,
some keywords, and a list of presentation formats (e.g., Flash, PDF, HTML);
this hardly supports a semantic search of LOs (and thus an efficient retrieval since
for example there may be thousands of LOs on Java) and even less the comparison of
these LOs (it is also necessary to note that the more the LOs relate to a great number
of objects, the more their comparison is difficult; to compare two concepts or two ideas
is possible because one of them can for example be a specialization, an argument or a
sub-task of the other but it is extremely rare that two articles or two documentations
can be connected by such relations).
Similarly, the modelings of the preferences and knowledge of the students are often very poor; for example: a keyword for each known LO (e.g., "Java") and a learning level for it (e.g., "advanced"). This does not permit a software to know what the student actually knows, as opposed a more fine-grained approach in which all the relations for which a student has been successfully tested on are recorded.
Similarly, the fact that knowledge is not modelled in a finer and more formal way limits the possibilities that each grid user (student, pedagogue, expert, researcher, employee, director, etc.) has for taking part in the development of the shared knowledge (this includes the LOs).
This project recognises the fact that a knowledge base (which may be general or specialized but which in any case must be detailed and formal) is the ideal tool for the retrieval, comparison, exploitation and re-use of knowledge, be it intended for learning, for decision-making, for the retrieval, comparison or evaluation of techniques (e.g., the retrieval and evaluation of the originality of research outputs), for collaboration or for decision-making. Although such KBs are very difficult to create manually (and even more automatically), there are some (very) long term projects the ambition of which is to create such KBs, for example the QED project which aims to create a KB of all the important mathematical facts, the OpenGalen project which aims to establish an ontology of medicine (but however not of all knowledge in medicine), and finally the "Digital Aristotle" the ultimate goal of which is to represent and be able to teach all established scientific knowledge. The goal of the project proposed here is not to create such KBs but to allow and encourage the "users" of semantic grids to create knowledge as precise and structured as possible, and hence also to improve knowledge exploitation. From now on, the term "users" is used for designating all knowledge providers and consumers (students, teachers or experts; users of one or more grids). The expression "as precise and structured as possible" recognizes that the input of "semi-formal" knowledge must be allowed for this approach to be sufficiently quick and then adopted by people (thus, the terms used for the concepts and the relations do not have to be formal, that is, declared) but 1) any sentence (formal or not) and any formal term must be connected to at least one other sentence or term by a formal relation (as alluded to in the previous paragraph, we consider that this structure of semantic network is a minimal structure to support knowledge comparison and sharing), 2) a lexical ontology including at least all common concepts (and relating them to all common words) must be proposed (Martin (2003b) proposes such an ontology but it requires many extensions and corrections for it to offer a support that is sufficient for knowledge representation and sharing), 3) this ontology must be sharable and be updatable by all the users (we consider that a set of partially redundant or inconsistent ontologies does not allow a sufficient comparison and sharing of knowledge because it is very difficult for a person - and a fortiori for a software - to inter-connect in a correct way ontologies independently developed from each other and without the guide of a large ontology).
This last point does not mean that the users must discuss and agree with each other in order to avoid lexical and semantic conflicts. Lexical conflicts are easily avoided by prefixing each formal term with the identifier of its creator, and semantic conflicts are avoided by the protocols proposed by (Martin, 2003). These protocols allow the users to have different "beliefs" but force the users to explicit the relations that exist between sentences that are contradictory or partially redundant. Examples of such relations are: "corrective_specialisation", "corrective_generalisation", "correction" and "example". The choice between contradictory beliefs does not have to be made (in a necessarily arbitrarily way) in the KB; it can be made later in the context of an application and by taking into account the relations which were set between the contradictory or redundant beliefs (for example, a simple strategy is to always select the most specialised beliefs, i.e., the most precise). Another simple strategy is to filter out beliefs coming from certain users, for example those having subscribed to certain types of newspapers (assuming that such information is stored). Finally, Martin & al. (2005) propose a method to evaluate the originality and popularity of certain beliefs, and then of their creators. This method is based on the argumention relations that are set between sentences (ideas, beliefs) and on votes by the users on the originality and popularity of these sentences. This method can be personalised by each user and thus constitutes a potentially effective means to filter the KB in order to see only the parts which a user wishes to see or use. Since this method attributes a score to each sentence and to each user, it also constitutes a way to encourage users to enter precise and original ideas. The above ideas were implemented by Martin (2003) and Martin & al. (2005) into a knowledge server (for reasons of facility). However, nothing seems to prevent the implementation and extension of these ideas into a peer-to-peer system or then into a semantic grid (actually, a system of replication of knowledge between peer servers is proposed by Martin & al. (2005)). Thus, this will be one of the axes of this project and the S-OGSA architecture will be exploited for that.
The two other axes of this project will be 1) to encourage the adoption of this approach by the production of LOs following the above cited criteria, and 2) to create applications exploiting the KB of a semantic grid and the results of other semantic grid related research to create applications allowing "ubiquitous learning" (as indicated above, the KB can be distributed and/or duplicated in the various machines of the grid). Although the term "learning" is used, the approach of this project actually intends to support any knowledge retrieval, comparison, sharing and update. Hence the title of the project.
Allison C., Cerri S.A., Ritrovato P. & Gaeta M. (2005). Services, Semantics and Standards: Elements of a Learning Grid Infrastructure. Applied Artificial Intelligence Journal (Vol. 19, No. 9-10, pp. 861-879), Dec. 2005.
De Roure D., Jennings N. & Shadbolt N. (2003). The Semantic Grid: A future e-Science infrastructure Grid Computing - Making the Global Infrastructure a Reality (Chichester, UK: John Wiley and Sons Ltd., pp. 437-470).
Gouardères G., Saber M., Nkambou R. & Yatchou R. (2005). The Grid-E-Card: Architecture to Share Collective Intelligence on the Grid. Applied Artificial Intelligence (Vol. 19, No. 9-10, pp. 1043-1073), Dec. 2005.
Martin Ph., Eboueya M., Jo J. & Uden L. (2006). Between too informal and too formal. Proceedings of KMO 2006, International Conference on Knowledge Management in Organizations (UM FERI; editors: M. Hericko, A. ZivKovic; pp. 38-87; ISBN 86-435-07809-6), Maribor, Slovenia, June 13-14, 2006.
Martin Ph., Blumenstein M. & Deer P. (2005). Toward cooperatively-built knowledge repositories. Proceedings of ICCS 2005, 13th International Conference on Conceptual Structures (Springer, LNAI 3596, pp. 411-424), Kassel, Germany, July 18-22, 2005.
Martin Ph. (2003b). Correction and Extension of WordNet 1.7. Proceedings of ICCS 2003, 11th International Conference on Conceptual Structures (Springer, LNAI 2746, pp. 160-173), Dresden, Germany, July 21-25, 2003.
Martin Ph. (2003). Knowledge Representation, Sharing and Retrieval on the Web. Chapter 12 of a book titled "Web Intelligence" (Springer; editors: N. Zhong, J. Liu, Y. Yao; pp. 263-297; ISBN 3-540-44384-3; Web Intelligence Consortium's book), January 2003.