An Information Visualisation Framework for Accessing ... - CiteSeerX

2 downloads 0 Views 2MB Size Report
{joris.klerkx, stefaan.ternier, michael.meire, katrien.verbert, erik.duval}@cs.kuleuven.ac.be. Abstract: In this paper, we discuss the use of Information Visualization ...
An Information Visualisation Framework for Accessing Learning Objects Repositories Joris Klerkx, Michael Meire, Stefaan Ternier, Katrien Verbert, Erik Duval Dept. Computerwetenschappen, Katholieke Universiteit Leuven Celestijnenlaan 200A, B-3001 Heverlee, Belgium {joris.klerkx, stefaan.ternier, michael.meire, katrien.verbert, erik.duval}@cs.kuleuven.ac.be

Abstract: In this paper, we discuss the use of Information Visualization techniques to improve the access to learning object repositories. We elaborate on a framework that was developed for this goal and on a prototype that was designed to enable users to access learning object repositories in a more flexible way than filling out electronic forms. In a second part, we present a case study in which we used the developed framework to visualize all the proceedings of the EdMedia. Users can use this application to find publications that were presented at the EdMedia conference over time. In the last part we try to elaborate on our ideas to visualize the internal components of learning objects and how we will try to visualize networks of interoperable learning object repositories.

Introduction We would like to discuss the ongoing research of the use of Information Visualization techniques to improve the access to learning object repositories (LORs) in our research-group. Information visualization is the use of computer-supported interactive visual representations of abstract data to amplify cognition (Card, Mackinlay, Shneiderman 1999). We believe that applying information visualization techniques can enable users to find learning objects in learning object repositories in a more flexible and effective way than filling out electronic forms, which is the way it is currently done in repositories like Ariadne, Edna, Merlot, Nime & Edusource. First of all, we would like to report on the development of a prototype framework which uses information visualization techniques and aims to enable users to find learning objects in a LOR in a more flexible and effective way. This framework was developed with the requirement to be as extensible as possible. In the second part of this paper, we describe a case study that was created to be a proof-of-context of this requirement. This case-study consisted of the visualization of the proceedings of all EdMedia-conferences. The last part elaborates on ongoing research taking our research and our prototype a step further.

Access to Learning Object Repositories In earlier work we described how we could use Information Visualization techniques to improve the access to learning object repositories (Klerkx, Duval, Meire 2004). In this work we presented the use of three information visualization techniques to visualize a complete LOR. These three techniques were tree-maps, hyperbolic trees and Venn-diagrams. This paper discusses how we took this work up to the next level and how we developed a framework and prototype which can be used to find learning resources in LORs. Basically the goal of our prototype is to enable an end user like a teacher or a student to zoom in on a relevant learning objects without requiring him or her to to go through a lengthy process of formulating complex search criteria, evaluating some of the results, refining the search criteria, etc (Duval, Hodgins 2003). We try to accomplish this goal by using Information Visualization techniques and by enabling end users to manipulate controls over the metadata, zoom in on potentially more relevant learning objects and continuously keep an overview of how additional search criteria restrict the remaining number of learning objects.

The prototype (Fig. 2) shows a screenshot of our prototype. The user interface of this prototype consists of three parts. The left part shows a visualization of the learning resources in the Ariadne Knowledge Pool System (KPS). In this figure, a treemap is used to visualize the semantic hierachical classification of the learning object. A tree-map is a visualization of hierarchical structure that makes 100% use of the available display space. It maps the complete hierarchy onto a rectangular region in a space-filling manner (Shneiderman, Johnson 1991). The semantic hierarchical classification is based on the ARIADNE-metadata, which is an application profile of the Learning Object Metadata (LOM). A path of taxons represents the classification of each learning object in the metadata. For example a possible semantic classification of a learning object about the Fibonnaci numbers is formed by the taxon-path that is illustrated in (Fig. 1). Each classification-taxon is represented in the visualization as a grey rectangle (Fig. 2). Science Type => Exact, Natural and Engineering Sciences Main Discipline => Informatics/Information Processing Sub Discipline => General/Sundry Main Concept => Complexity of Algorithms

Figure 1: Classification of a learning object about the Fibonacci Numbers In our earlier work (Klerkx, Duval, Meire 2004) we described the advantages of using such a tree-map visualization to visualize the LOR. Among those advantages are the insight they provide in the hierarchical classification and the overview they provide of the entire LOR.

Figure 2: Screenshot of the prototype. The left part consists of a tree-map visualization of all learning objects in the Ariadne LOR; the right part consists of some metadata controls and an information panel.

In the tree-map view of (Fig. 2), no actual learning objects are visualized as such by default. Only the numbers of learning objects that match a specific classification taxonpath are visualized. We say that a learning object matches a specific classification taxonpath if two conditions are fulfilled: 1. 2.

The classification taxonpath of the visualization is the same as the classification taxonpath that can be found in the metadata of the resource. The learning object matches the filter criteria that can be found in the right part of the user interface.

With these filters, the user can quickly change the number of matching learning objects by changing the range sliders. When the sliders are adjusted, the numbers of matching learning objects will also be automatically changed. The more resources are matching to a specific classification taxonpath, the more space the classification-rectangle gets assigned in the visualization, which can be seen in the figure. The number of actual matching learning objects is also displayed in the rectangle as a number so the user has a clear overview of the remaining learning objects. The green color in the above figure was chosen because we wanted the matching resources to be strikingly visible. The green color is hard-wired into the human brain as a primary (Ware 2004) and was therefore chosen for this purpose. The user can also navigate in the visualization itself by clicking on the rectangles. If he clicks on a rectangle that represents a classification, the application will zoom in onto that rectangle. When he clicks on a rectangle that represents the matching learning objects of a classification taxon-path, the application will show specific information of all the learning objects that matches that rectangle in the right lower part of the interface. It will also zoom in again and show the different learning objects as is shown in (Fig. 3). The user can than again zoom in by clicking on the learning object that seems interesting to him or her. More information on the learning object will then again be shown in the right lower part of the interface. He can access this resource by clicking on a link that will bring him to the full description in the Ariadne KPS web interface where he can also download it.

Figure 3: A screenshot that shows 13 matching learning objects The framework The framework was developed so that it would be as open and extensible as possible. We wanted to be able to rapidly add new visualization-techniques at the one hand and new kinds of components that are described by possibly other metadata-schemes at the other hand. This was a requirement because we wanted to be able to reuse the framework in other contexts. It should for instance be possible to visualize a digital library of music-tracks, with a minimal amount of work. As a proof-of-concept that we can easily extend our prototype with different visualization-techniques, we added a hyperbolic tree visualization to visualize the LOR, which can be seen in (Fig.4).

The current framework is coupled with the KPS-Client of the Ariadne system[1] to visualize the Ariadne KPS. By doing this, we get an additional advantage, namely that we can not only visualize learning objects of the Ariadne KPS, but we will also be able to visualize the learning objects of other LORs like Merlot[2] because those LORs are interoperable by using the Simple Query Interface (SQI).

Figure 4: A screenshot of the prototype that uses a hyperbolic tree instead of a tree-map Evaluation We strongly believe in rapid-prototype developing as it allows the developers to hold examples of their concepts for early visualization, verification, iteration, and optimisation. Since the earlier work (Klerkx, Duval, Meire 2004) we conducted some rapid user-tests of the use of different information visualization techniques we investigated in the context of getting resources out of a LOR. The results gave us enough reason to push forward with our research and the development of our prototype. A more extended evaluation of the current prototype with extensive user-tests is ongoing and will be completed within the near future.

Case study: visualising the EdMedia proceedings As a proof-of-concept of that we can easily use our framework in another context, we visualized the proceedings of the EdMedia conferences [3] [4] through the years. With this visualization, we can find publications in a more interactive way. A screenshot of the visualization of the EdMedia proceedings can be found in (Fig. 5) and (Fig. 6). We chose to classify the publications by year. As an example, a user searching for a paper that covers learning objects, immediately gets an overview of the publications on that topic over time. If he or she wants to exclude earlier papers, the user can change the sliders in the right part of the interface and the number of matching publications will change. If he or she wants more information on matching papers, he or she can click and the matching paper will show information like the title, author, abstract and the document-handle in the resultsinformation page. We created this visualization based on the metadata of the publications that was created by using an automatic metadata generation framework (Cardinaels, Meire, Duval 2005), which is made available as an indexing web service. The idea behind this framework is to combine metadata, generated by different sources into 1 metadata [1]

http://rubens.cs.kuleuven.ac.be/ariadneDoc http://www.merlot.org [3] http://www.aace.org/conf/edmedia [4] http://www.aace.org/dl [2]

instance. In our case, we wrote an extension or plugin for the framework, dealing particularly with the AACE digital library and more specifically with the EdMedia proceedings in that digital library. This piece will make use of the information we have about the content to generate metadata like the title, the authors, the document handle and the abstract of each publication. The advantage of using this framework is that it already contains components that generate other parts of a metadata instance, like the language of the document. The result of calling the IndexingService will therefore be an automatically generated (rich) metadata instance that contains as much metadata as possible.

Figure 5: A visualization of all matching proceedings of the EdMedia conference, classified by year

Further research First of all, we will work on functionalities that are helpful to users like e.g. the ability to mark and annotate learning resources so that they can use these marked and annotated resources for future use. He should be able to do this for himself and other users. We will also carefully complete the usability evaluation that we described earlier and change the prototype according to these results. Secondly, we will proceed with our research on visualising the different internal components of the learning objects and on visualising a network of repositories. The reasons we want to pursue these ideas are described in the next sections. Visualising the components of learning objects Until now, we visualized the different learning objects, solely based on the metadata that describes them. The next step is to go a step further and to visualize the internal components of these objects so that end users can get a good overview of the learning resource itself. For this purpose we can use the research on ontology-based learning content repurposing that has been done in our research team. This ontology is a solid basis for an architecture that will enable on-the-fly access to learning object components and that will facilitate repurposing these components (Verbert, et al 2005a). At the moment a framework, based on this ontology, is already developed to aggregate and disaggregate learning object components, more specifically components of slide presentations (Verbert, et al 2005b). We will do more research how we can visualize the internal structure of the learning objects, based on this framework. This work could then result in e.g. an extra zoomlevel in our prototype that is described above. This

extra level could show the internal components of the learning resource and also the interrelationships between those components.

Figure 6: A visualization of all proceedings of the EdMedia conference, classified by year. The matching proceedings are visualized with a green color. It would also be interesting if we could visualize which components are used in which context. An image of the pyramids could e.g. be used in the context of a history course at a university but it could also be used in the different context like a brochure that describes Egypt. This information could e.g. be used in the following scenario: when a teacher is creating a new learning resource, he would like the technical support that would automatically search LORs for a particular image, graphic, paragraph, etc. The teacher can then select the components he wants to repurpose in the resource he is creating (Verbert, et al 2005b). This could still be a complex task because of the number of automatically proposed components. If we are able to create a visualization that gives an insight of the usage of the proposed components in different contexts, the teacher should be able to rapidly filter out uninteresting components. Visualising a network of repositories Apart from visualizing the content of particular repositories, visualising a network of repositories could provide very useful information. The recently launched GLOBE consortium aims to unlock the ‘deep’ web and promises its members (Ariadne, Edna, Merlot, Nime & Edusource) to do federated search in each others repositories (Duval et al. 2004). Federated search allows client applications to search in multiple repositories at once, taking advantage of the metadata that these repositories maintain. Not only can we opt to visualize the content of these repositories, but we can also look into the possibilities of visualising and modelling a landscape of interoperable LORs. These landscapes could provide users with helpful information like e.g. the physical location of a learning resource in the internet IP world which could be used to get an idea how fast or slow it will take to get access to it in terms of bandwidth. A different kind of landscape could give information of the geographical location of the repository. By knowing this geographical location, it is possible to get an idea how easy it would be to directly contact the creator of a learning resource and possibly cooperate with him or her.

Visualising a landscape of networked repositories could also provide insight in the actual content of the federated repositories. As some repositories are used within a context, it makes sense to visualize this context. The EdMedia digital library e.g. will only store papers on educational multimedia, hypermedia & telecommunications while a learning object repository that acts as a storage layer for a Learning Management System, will probably cover more heterogeneous content. We can imagine that the user would like to see this context. We can also imagine that a user wants to see which repositories contain large numbers of learning resources on a specific topic like economy, and more specifically those repositories that would actually contain components that have been repurposed for a number of times within a context that matches the one of the user.

Conclusions Currently, users are able to search for learning objects in LORs by filling out electronic forms that enables them to compose boolearn combinations of search criteria. In this paper we took up the work we started in (Klerkx, Duval, Meire 2004) to investigate how we can use information visualization to improve this kind of access. More specifically, we discussed a prototype framework that was developed to enable users to manipulate controls over the metadata, zoom in on potentially more relevant learning objects and continuously keep an overview of how additional search criteria restrict the remaining number of learning objects. We also described a case-study that was meant as a proof-of-concept that this framework can be used in different contexts. This case-study visualizes all the proceedings of the Ed-Media conferences and can e.g be used to follow the research on a particular topic over time. We concluded by discussing the direction of our research. More specifically we will do more research on visualising the internal components of the learning objects and on visualising a network of interoperable LORs. All these efforts will hopefully result in a more effective and flexible access to learning resources in learning object repositories.

References Card, S., Mackinlay, J. D., Shneiderman, B. (1999). Readings in Information Visualisation, using vision to think, Morgan Kaufmann Publishers, Inc. Klerkx, J., Duval, E., Meire, M. (2004). Using Information Visualisation for Accessing Learning Object Repositories, In Proceedings of IV04 8th International Conference on Information Visualisation, London, England, 465-470. Duval, E., Hodgins, W. (2003). A LOM research agenda, In Proceedings WWW2003 12th International Conference on World Wide Web, Budapest, Hungary, 1-9. Shneiderman, B., Johnson, B. (1991). Tree-maps: A space-Filling Approach to the Visualization of Hierarchical Information Structures, Proceedings of IEEE Information Visualization ’91, pp 175-282 Ware, C. (2004). Information Visualisation, perception for design, second edition, Elsevier, Morgan Kaufmann Publishers, Inc. Cardinaels, K., Meire, M., Duval, E. (2005) Automating Metadata Generation: the Simple Indexing Interface, (submitted). Verbert, K. et al, (2005) Ontology-based Learning Content Repurposing, (submitted). Verbert, K. et al, (2005b) Towards a Global Component Architecture for Learning Objects: a Slide Presentation Framework, (submitted) Duval, E. et al, (2004) Press-release: http://rubens.cs.kuleuven.ac.be:8989/mt/blogs/ErikLog/archives/000665.html Acknowledgements We gratefully acknowledge the financial support of the K.U. Leuven Research Fund, in the context of the BALO project on “Basic research on Learning Objects”.