THE VISUALISATION AND EXPLORATION OF

A CURRICULUM KNOWLEDGEBASE

Sebastian Richards and Hilary Dexter

Manchester Medical School, University of Manchester

Keywords: Visualisation, Knowledgebase, Curriculum, Exploration, Navigation.

Abstract: This paper discusses the difficulties associated with the visualisation and navigation of complex, multi-

faceted knowledgebases. The issues are analysed through a case study of the Manchester Medical School

curriculum, with the intention of developing an innovative visualisation tool to enhance the experience of

navigation through such large sets of data. The approach taken to provide a solution involves the use of a

metaphorical representation of the data schema, designed to reduce visual complexity and increase human

usability. User feedback uncovered the need to create a ‘trail’ of related concepts through the

knowledgebase, and provision for such navigation techniques is present in the developed tool. The concept

of domain-specific structural customisation was investigated and led to the suggestion that modification of a

schema towards a more logical and human-friendly abstraction can improve a user’s understanding and

engagement with data.

1 INTRODUCTION

During the development of a tool to aid a team to

maintain and enhance an undergraduate curriculum,

an issue arose concerning the visualisation of the

rich and complex knowledgebase which

underpinned the tool. For the team to engage with

curriculum maintenance and enhancement they

needed to look at large sets of interrelated data

items, retrieved from the knowledgebase. This

presented both conceptual and technical challenges.

The curriculum domain is complex in terms of the

number of types of data and their interrelationships,

and large in terms of the number of instances of

some of the types, making convenient visualisation

of useful selections a pressing requirement. This

paper describes an investigation into the

development of a visualisation tool for such a

knowledgebase. Although developed initially for the

curriculum domain, it is generally applicable to any

area.

The aim of the investigation was to find a way to

avoid the common problems associated with

viewing large, complex sets of data and to produce a

visualisation that depicts an easily digestible view of

a multifaceted structure. The visualisation had to

afford a better understanding of the interrelated

concepts, thereby allowing the user to focus on the

essentials of the amendments to the curriculum. The

case study presented here is the updating and

enhancement of the undergraduate medical

curriculum in Manchester Medical School. The

people involved in this felt that such a tool would

assist in the immediate task of curriculum

enhancement and would also serve to develop the

usability and embedding of the knowledgebase

technology proposed for use general by the core

educators group (Dexter and Davies, 2009).

The approach taken was to find a simple and

intuitive visual metaphor and useful ways of

grouping data. We present here the key ideas behind

the simplification of the visualisation problem, and

address some of the human-computer interaction

(HCI) issues inherent to communicating large data

sets. The visualisation tool, which is under ongoing

development, aims to deliver increased usability and

provide a key set of features:

• To simplify the challenge of gaining a

meaningful understanding of the data

• To help a user easily realise relationships

between concepts

• To discover trails between linked concepts

through the knowledgebase.

168

Richards S. and Dexter H. (2010).

THE VISUALISATION AND EXPLORATION OF A CURRICULUM KNOWLEDGEBASE.

In Proceedings of the 12th International Conference on Enterprise Information Systems - Human-Computer Interaction, pages 168-173

DOI: 10.5220/0002907301680173

 SciTePress

2 KNOWLEDGEBASE

EXPLORATION

Due to the complex nature of a curriculum, the

underlying data structure of the representational

model makes manual exploration difficult. A

common issue is that concepts from the knowledge

domain frequently will have large numbers of many-

to-many interrelationships, giving rise to a high data

density. As a consequence, data visibility diminishes

and an overwhelming number of connections can

prevent a user from easily grasping what the data

actually represents. When viewing such a structure

as text, it is often hard to mentally visualise the data

and gain an understanding into how concepts may be

related (Cleveland, 1993). This difficulty escalates

rapidly with increasing size of a knowledgebase.

The limitations associated with viewing a

knowledgebase at a textual level are often overcome

with the use of a visualisation tool. Frequently,

however, these tools become unusable when the

number of concepts in view becomes too large,

typically maintaining a high visual complexity that

provides both unnecessary and unwanted detail. The

overall effect is a dramatic decrease in data visibility

that ultimately fails to provide the user with any

insight or meaning. The impracticality of such an

approach to visualisation for detailed work, rather

than simply gaining an overall impression, is shown

in Figure 1. The example is taken from (What is an

ontology?, n.d.), and demonstrates a structure of

hierarchical classes populated with many interrelated

individuals. This is a typical ontological structure,

and is comparable to that of the knowledgebase used

in our work with the curriculum knowledgebase,

CRAMPON (Dexter and Davies, 2009). Although

the visualisation offers a depiction of the overall

structure of the ontology, it is not aligned with the

needs that our solution aims to support.

3 REDUCING VISUAL

COMPLEXITY

In the aforementioned curriculum modelling

research (Dexter and Davies, 2009), it was proposed

that a potential method for reducing the apparent

structural complexity of a knowledgebase could be

to use the proven real-world solution of an

underground railway map; this type of adaptation is

used internationally to simplify the ever complex

and multi-dimensional physical constructions

associated with such structures. In the case of the

Figure 1: Typical ontology visualisation.

London Underground map, the introduction of an a-

geographical map has been found to provide an

embedded ‘city-user’ interface, where human

interaction continues far beyond the scope of railway

navigation (Vertesi, 2008). In an instance where a

similar approach was used to create a visualisation

of a complex business process model, feedback was

generally positive (Burkhard and Meier, 2004).

In the proposed map, the types or ‘classes’ of

concepts would be represented as train stations and

the rail lines between them would represent the

relationships between concepts. The user would

select a ‘route’ through the knowledgebase by

choosing sequential stations that were connected by

rail lines, in a manner analogous to selecting routes

across a city. This simplified representation of the

knowledgebase can not only dramatically reduce the

number of elements on view, but also replace an

unwieldy data structure with a symbolic and

recognizable metaphorical one. Although the use of

grouping is found in existing visualisation tools,

conventional approaches typically offer only

temporary benefit whilst a group is collapsed; in

short, it allows for only one level of abstraction to be

used at any one time. This restriction forces a user to

view the knowledgebase either in its entirety or at

group/sub-group level, preventing inter-group

relationships between instances from being shown.

In comparison, the route-mapping approach

enables a user to see the entire knowledgebase

structure through the selection of train stations on

show, with the possible relationships between

instance classes shown by the rail lines between the

stations. Once a user has identified the stations and

routes relevant to the data they are concerned with,

THE VISUALISATION AND EXPLORATION OF A CURRICULUM KNOWLEDGEBASE

169

those can be selected and the relevant instances

listed. This approach offers the user a selection of

the data they may wish to view before it is shown,

rather than displaying all of the data immediately.

By introducing this intermediate stage of filtering,

the knowledgebase can be displayed and navigated

in detail, while avoiding the common pitfall of over-

crowding.

To test the usability and potential of the rail map

metaphor for finding ways though the knowledge

domain, a number of workshops were conducted

with a range of participants, not solely limited to the

target audience of the curriculum teams. In these

workshops, several knowledgebase exploration

interfaces were offered for consideration: a drag-

and-drop user interface (UI), a matrix lookup UI and

the route mapping UI. Rather than displaying the

structure of the knowledgebase directly, the drag-

and-drop interface showed the classes or ‘stations’

and allowed a user to select a sub-set. Although the

interface was not a conventional visualisation of the

data, it offered an intuitive method for creating a

complex query based on the selection of only

relevant classes and relationships. One flaw with this

process however, is the requirement for an

understanding of the knowledgebase structure,

which restricts its use to those with prior experience.

The matrix interface offered the similar approach

of selecting a base class on which to find instances,

as well as selecting the set of classes that each

instance should be related to. This method suffered

from the same pitfalls as the former, by requiring at

least some understanding of the knowledgebase

structure beforehand. From the workshops it became

clear that an unfamiliar user would greatly benefit

from the structure being presented in its entirety. It

was also apparent that users felt positive about

finding useful routes, but they needed a more robust

solution than that offered by the rail map. Since the

rail map would display all of the relationship routes

available between any of the stations at any given

time, it was prone to the same over-complex

appearance of the typical visualisation techniques; a

revision of the metaphor was required.

4 EXTENDING THE METAPHOR

Analysis of the feedback from the workshops led to

a revision of the underground railway metaphor to

something that could offer greater flexibility: an

airline’s routes to a set of airports. In this visual

metaphor, airports represent the knowledgebase

classes, and flights represent the associations

between the instances of the classes located at each

airport. This new approach aims to follow a style of

route-mapping commonly found on airline websites;

the example tool by Easy Jet (EasyJet Route Map,

n.d.) shows all of the airports serviced by their

flights, along with the routes from a particular

airport once selected. The characteristic of

displaying only the routes from one airport at any

given time greatly assists in reducing crowding. Had

the entire set of possible flight paths been shown, the

visualisation would have been unclear and unusable.

By using a similar technique to the economy

airlines, a suitable level of visualisation detail was

reached, as shown in Figure 2.

Figure 2: Visualisation using ‘airport’ metaphor.

Although a setting for showing all the routes does

exist within the tool, it is provided only to offer an

overview into the connectedness of the

knowledgebase structure. The instance count at each

airport is shown after its title. Once an airport has

been selected and its routes shown, the numbers at

each of the related airports will change to reflect the

number of related instances. The instances for the

selected airport are displayed in a grid below the

visualisation with any relevant properties. Keyword

filtering can be applied at any airport, allowing a

user to view only the routes from the resulting

instances. The related instances can then be viewed

by selecting the relevant target airport.

In order to make provision for the hierarchical

structure found in many knowledgebases, the notion

of an airport terminal is used to model sub-classes.

Due to the potential over-crowding of these

terminals being drawn on the map, they are shown in

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

170

Figure 3: Complete visualisation interface.

a separate panel in the interface. When viewing the

instances of an airport on the grid, the terminals are

shown as collapsible groups. Figure 3 shows the

visualisation tool in its entirety. The right-hand side

panel shows the keyword filter and terminals

control, while the lower section shows the instance

browser. If the user wishes to start a journey with an

instance (or set thereof), they can simply ‘tick’ the

relevant box in the instance browser.

To enhance the visual realism of the metaphor a

number of ‘islands’ are shown on the map, providing

a more realistic feel to the map. Also, the routes

between each airport are drawn as Bezier curves.

USER WORKFLOW

A common feature among existing visualisation

tools is the lack of focus for their use (Katifori et al.,

2007). Although maximum flexibility is maintained

through lack of rigid usage specifications, providing

a user with a set of pre-determined workflows can

increase the simplicity of using such a tool. As part

of the workshops used to evaluate the route-mapping

metaphor, a number of workflow patterns were

determined. By establishing the desired workflows

of the target audience, it is possible to produce a

more relevant and usable tool.

Although the needs of individual users are likely

to vary significantly, a commonly identified

requirement was the discovery and realisation of

inter-conceptual relationships. Rather than leaving

the user to attempt the task using an ‘all-or-nothing’

approach, the route-mapping metaphor permits all

levels of abstraction to be used simultaneously. This

is achieved by providing a high-level view of all

airports and a set of instances with relationships

between them (as flights) simultaneously; both the

THE VISUALISATION AND EXPLORATION OF A CURRICULUM KNOWLEDGEBASE

171

individual elements and entire dataset are

observable, along with the relationships or ‘routes’

between them. From this position, a user is able to

assimilate easily information on both the structure

and relatedness of the data within a knowledgebase

at multiple levels.

To further the capabilities of information

discovery, the ability to build a ‘journey’ through a

series of airports makes provision for another crucial

workflow pattern. By building up a trail of inter-

related concepts across a knowledgebase, a user can

generate a view that depicts how even the most

distant elements are connected, thus gaining further

insight into the real meaning of the data. In practice,

this can enable a user to not only see how and where

data is related, but also where it is not. When used in

the domain of a curriculum, this functionality can

assist in highlighting areas of teaching that perhaps

have not been given sufficient coverage, or perhaps

identified as relevant to another part of the

curriculum.

6 STRUCTURAL

MODIFICATION

The application of a visual metaphor to the structure

of a knowledgebase effectively removes any direct

contact between the structure and user. The elevated

mediated view removes the invariably complicated

appearance of data (as demonstrated by Figure 1),

thus introducing an abstract version of the data

model. This intermediate layer can provide a

valuable opportunity to implement domain-specific

logic to the way the visualisation supports

interaction, offering yet another mechanism for

simplification and user guidance. In the case of the

Medical School curriculum, this allowed for parts of

the curriculum not relevant for visualisation to be

omitted. Similarly, the tool was able to display a

modified knowledgebase structure, facilitating the

emulation of a hierarchical structure that existed

logically rather than within the database schema. For

the curriculum visualisation, this was used

specifically to enforce parent-child relationships

where the children were logical sub-components

rather than literal sub-types.

7 PERFORMANCE & USABILITY

Although the original intentions of using an

ontology-based knowledgebase proved both flexible

and convenient, the addition of approximately 7000

elements to the model saw the speed of data retrieval

dramatically decrease to an unacceptable level. This

issue prompted a move away from an ontology

based knowledgebase to a model-driven-architecture

framework, Sculptor (Sculptor, n.d.). This

technology provides a set of re-usable database

interfaces, generated automatically from a domain-

specific representation of the data model. Any

revision to the structure simply requires that the

model is modified and data is migrated following re-

generation.

One crucial downside to this move is the

inability for the visualisation tool to dynamically

determine the structure and possible associations

that exist within the knowledgebase. Instead, the

visualisation tool requires a version of the

knowledgebase schema that can be used to render

the desired model and specify the actions and

relationships for each class. A possible future

development could see a specific version of the

schema created synchronously with the model-

driven generation of the database infrastructure; this

automated step would simply require modification

of the Sculptor generation chain. It is worth noting

that this issue is the result of the choice of database

architecture; the visualisation tool itself is not tied to

any back-end architecture.

8 SUMMARY AND NEXT STEPS

The tool developed during this research provides a

set of solutions for issues commonly experienced in

knowledgebase visualisation. The route-mapping

metaphor offers a comprehensive representation of

raw data structures that is used effectively in

reducing visual complexity. Coupled with the

provision for building a ‘journey’ or ‘trail’ through

the structure, an overall increase in data visibility

and intention is delivered.

Further evaluation of the success and potential of

the tool is planned. The visual metaphor will be

assessed against the metrics and guidelines set out

by (Shneiderman, 1996) and (Burkhard, 2005).

Future use of the tool both within and outside of the

medical domain will provide real-world feedback

from a range of users, outlining possible limitations

and determining potential improvements. In order to

produce a quantifiable assessment of the tool, a set

of activities will be derived to compare the speed of

information discovery to that of an alternative,

rudimentary presentation of the data (e.g. tree

diagram).

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

172

Although a decision to avoid the conventional use of

clustering was made early on, one logical

progression for our visualisation is the ability to

‘zoom in’ on an airport and view its terminals

graphically. This differs from the conventional

approach of displaying an instance graph once a

group is expanded; the suggested development

would only be used to represent hierarchical

structures, ensuring expansion remained on the class

(or airport/terminal) level.

The ability to discover where data may be

unrelated has been beneficial during use in

curriculum exploration. An obvious extension of

functionality would be the provision of a facility to

automatically determine orphan instances; that is,

where an instance of any given class that has no

relationships to other instances where permitted by

the schema. This would provide a unique view of the

data in a way that is atypical of a visualisation tool.

As visualisation of data often provides a useful

output of information, the ability to generate a report

of findings from such a tool would be greatly

beneficial to users performing manual data analysis.

A mechanism for providing such functionality could

perhaps make use of a drag-and-drop interface,

where any object to be included in the report could

simply be dropped into a ‘report bucket’.

Another potential addition is that of a dynamic

context control, in the form of a drop-down box.

This would allow the visualisation to be given a

specific context to display, for example only data

from a particular year. The context would effectively

be a top-level filter on the data, where a

knowledgebase contained a relevant set of instances

that possessed qualitative relationships with

instances in other classes.

ACKNOWLEDGEMENTS

Thanks to Tim Cappelli (Project Manager) for

proposing the visual metaphors and designing the

user workshops.

The CRAMPON project was funded by JISC

REFERENCES

Burkhard, R. A. (2005) Towards a Framework and a

Model for Knowledge Visualization: Synergies

Between Information and Knowledge Visualization,

Berlin/Heidelberg: Springer.

Joint Information Systems Committee.

Burkhard, R. A. and Meier, M. (2004) 'Tube Map:

Evaluation of a Visual Metaphor for Interfunctional

Communication of Complex Projects', Proceedings of

I-KNOW.

Cleveland, W. S. (1993) Visualizing Data, Hobart Press.

Dexter, H. and Davies, I. (2009) 'An ontology-based

curriculum knowledgebase for managing complexity

and change', Ninth IEEE International Conference on

Advanced Learning Technologies.

EasyJet Route Map, [Online], Available: http://

www.easyjet.com/asp/EN/routemap/ [18 Jan 2010].

Katifori, A., Halatsis, C., Lepouras, G., Vassilakis, C. and

Giannopoulou, E. (2007) 'Ontology visualization

methods - a survey', ACM Computing Surveys, vol. 39,

no. 4.

Sculptor, [Online], Available: http://fornax.itemis.de/

confluence/display/fornax/Sculptor+(CSC) [12 Jan

2010].

Shneiderman, B. (1996) 'The Eyes Have It: A Task by

Data Type Taxonomy for Information Visualizations',

Proceedings of the IEEE Symposium on Visual

Languages, pp. 336-343.

Vertesi, J. (2008) 'Mind The Gap: The ‘Tube Map’ as

London’s User Interface', Social Studies of Science,

vol. 98, no. 1, pp. 7-33.

What is an ontology?, [Online], Available: http://

lsdis.cs.uga.edu/projects/glycomics/report/Report2006

.html [3 Jan 2010].

THE VISUALISATION AND EXPLORATION OF A CURRICULUM KNOWLEDGEBASE

173