|
Case Study:
EC-Funded Commission Creates Interactive Web-based "SmartDocs"
Professor
Mikael Jern
ITN, Linkoping University, Sweden and Advanced Visual Systems, Denmark
Abstract
The contents
in Web documents are normally restricted to static items such as
text, imagery and animations. SmartDoc has developed Collaboratories
(Web application components) that incorporate not only text but
also the entire interactive data visualisation and navigation process
into a Web document, allowing users and project teams to collaborate
and share data, visualization parameters, information and insight
while distributed over standard or mobile Internet, using intuitive
visual navigation techniques. In other words, publishing a text
document on the Web is only half the story. The other half is enabling
others to interact with the published result and gain insight into
context that's meaningful.
SmartDoc is
a research project jointly funded by the EC Commission and focused
on embedded Collaboratories that give the reader full access to
any discovery and insight, data navigation tools and underlying
data. Visual data navigation is provided through interactive 2D
and 3D Web-based visualisation components with a small footprint.
The "discovery" is described in one or several snapshots providing
the history of the visualisation process. These snapshots are a
copy of the component's state at the time when the snapshot was
taken and allow the user to further interact from the state when
the visualisation was snapped. They can be included as an image
for printing the document. The underlying data or spreadsheet is
either embedded in the document or accessed through a hyperlink.
"What are users looking for?" is the key question guiding a SmartDoc
process.
The Collaboratories
(Application Components) are based an multi-layer visualisation
component architecture (figure 2) with a small
footprint suitable for Web distribution and therefore scalable and
customisable to any level of expertise. A "SmartViewer" client-side
plug-in, responsible for interactivity and graphics rendering, has
been developed and will be distributed as "freeware" to allow free
distribution of a SmartDoc on the network. Integration and assessment
of application component sharing through Web documents and a network
infrastructure based on the component industry standards providing
real-time data interactivity, reducing the load on the network and
with zero administration client deployment.
SmartDoc scales
to accommodate massive amounts of data presented in a visual format,
allows full real-time interaction with on-screen presentations,
and gives users an unprecedented level of high-quality visual presentation.
Our integration of visualisation and data analysis through an atomic
component architecture combined with special data reduction components
and fast scene tree rendering by the SmartViewer enables the visual
data navigation of large data sets.
Introduction
to SmartDoc
The SmartDoc
visual data navigation components allow the author to embed not
only the content (text), but also the entire visual data navigation
scenario in any electronic document. Upon receipt of the SmartDoc,
the recipients can immediately share the author's insight and zoom-in
on specific results that affect their activities. The person viewing
the report has full access to any discovery (insight), data navigation
tools and underlying data. The "discovery" is described in one or
several snapshots ("bookmarks") also embedded in the report. These
snapshot are a copy of the control's state at the time when the
snapshot was taken and allow the user to further interact from the
state when the visualization was snapped. You can visualize multivariate
data and derive a deeper understanding of compound properties through
data correlation and multiple display methods of the same data source.
You can filter unwanted values using range sliders to expose areas
of greatest importance. Isolating data of interest clears the view
even further and provides close-ups of the results.
 |
|
Figure
1: The SmartDoc represents an interactive document with embedded
Text (context), Visual Data Navigation component, Data, Image
for communication and the Status of the exploration process.
|
An example data
might be a set of multivariate data perhaps related to product formulations,
consumer behavioural studies, or a set of measurement attributes.
Typically a standard report on such data might start off discussing
the data in general terms and then continue with the outputs of
some statistical or modelling studies to extract information from
the data and then draw conclusions perhaps coupling in additional
related information. Conceptually the interrelations within this
class of dataset and between the raw data and modelling process
are hard to imagine, making it difficult to extract the "learnings"
from the particular problem.
Instead imagine
a smart document that takes the reader through the problem via a
set of interactive visualisations coupled to the raw data. The reader
can not only read the report but if required do parts of the analysis
from within the familiar document interface, starting with exploratory
visualisation of the raw data through driving the modelling/statistical
process through assimilation of the final results.
By doing this
the reader:
- Will be able
to explore the raw data hopefully leading to an understanding
of why a particular analysis approach was subsequently taken.
- Get a much
more intuitive feeling for the trade-offs in the modelling approach
and therefore the validity of the results.
- Will take
in the substance of the report in a very natural way.
Upon downloading
of a SmartDoc to the desktop, the recipients can view text and images
but also share the author's visual insight and analysis and zoom-in
on specific results that affect their activities. You can rearrange
multivariate data to uncover patterns and trends that may go unnoticed
in lengthy paper reports. A SmartDoc is based on embedded Collaboratories
that give the reader full access to any discovery (insight), data
navigation tools and underlying data (figure 1).
Visual data navigation is provided through interactive 2D and 3D
Web-based visualisation components with a small footprint. The "discovery"
is described in one or several snapshots providing the history of
the visualisation process. These snapshot are a copy of the control's
state at the time when the snapshot was taken and allow the user
to further interact from the state when the visualisation was snapped.
The underlying data or spreadsheet is either embedded in the document
or accessed through a hyperlink.
 |
| Figure
2: SmartDoc component multi-layer user abstraction model and
system architecture |
The SmartDoc
EC funded project addresses the concept of "Information Visualisation"
(InfViz) through its support of:
- Visual data
mining techniques
- Visualisation
of multidimensional data · Interdisciplinary focus - visualisation,
data mining, VUI, data analysis, clustering
- Information
visualisation techniques for large data sets
- Integration
of visualisation techniques in information and knowledge management
systems
- Combination
and integration with non-visual data mining techniques
- Visual data
analysis and exploration for
- Knowledge
management
- Consumer
analysis / Marketing
- Biochemical
and biomedical analysis
- Internet
/ E-Business
- Emphasis
on application-oriented projects
 |
| Figure
3: Visualization of abstract data. Five columns (MPG, HORSEWPOWER,
…) in Excel representing a "car model data set" are mapped onto
a3D sphere graph. Multiple 2D and 3D views of the same dataset
can often help the understanding of trends and patterns in a
multivariate data set. As we become more analytically oriented
and begin to think multi-dimensionally, these 3D techniques
will become a necessary tool to deal with massive amounts of
complex abstract data. |
Data Model
Spreadsheets
have helped us organize our information into two-dimensional matrices
of rows and columns. Our data model incorporates multi-dimensional
data without any limitation on the number of dimensions supported.
Analysis tools slice and pivot among the dimensions to display the
data in any number of forms. The InfViz tools enable the user to
pivot among the dimensions interactively.
There are a
number of reasons for our choice of the Excel spreadsheet as the
data model:
- While we
might like to think that data is managed in properly designed
and maintained databases, the amount of data simply stored in
Excel spreadsheets, accessible on the users desktop is truly vast.
Tools that address directly the problem of exploring this data
will be of general interest.
- While data
will be drawn from a huge range of applications and problem domains,
ranging from finance, IT and business management, to manufacturing,
research and development, there are often some common issues,
for example the data may be highly multivariate and/or time dependant.
The opportunity therefore exists to develop a tool based around
a particular set of visualisation techniques optimised for a particular
class of data (multivariate) which never the less has wide applicability.
Further because the basket of required visualisation techniques
is therefore well defined, we can focus on exploring the value
of the COM approach in delivering and demonstrating the performance
required.
- While the
graphical techniques within Excel are improving they still remain
limited to simple presentational visualisations, there are no
interactive exploratory visualisation tools, the data navigator
can therefore be thought of as a tool for significantly expanding
the functionality for exploring multivariate data sets available
to the Excel user.
- Excel itself
is a component-based application.
3D InfViz provides
several methods of viewing multiple dimensions in the same display.
One way is to map the abstract data to a virtual object such as
a 3D sphere graph. Five columns in an Excel spreadsheet are mapped
onto X, Y, Z, Size and Colour (figure 3). Effective
users of this InfViz technique, however, must be analytical and
multidimensional thinkers. While it seems intuitive for analytical
thinkers to assess data by multiple dimensions not everyone thinks
that way. When information is organized dimensionally as in the
3D sphere graph, the potential for understanding it grows exponentially
and particular when combined with 2D views (figure
4).
The 3D sphere
graph shows how various car model attributes such as mpg, horsepower,
weight, acceleration, number of cylinders etc, correlate to each
other. The 3D spheres reveals, for example, that Japanese cars use
less petrol (MPG) than American cars. Is this due to number of cylinders
or weight? How is the acceleration of a car model related to its
weight etc? The colleagues that you involve in this analysis will
want to ask their own questions of the data. They'll want to start
with the same perspectives that sparked the questions, then apply
their own expertise and experience to the analysis by branching
off into their own exploration of the data. A useful way to derive
insights from your analysis is to create an intuitive map to saved
views of the data, and then make this discovery available to colleagues.
You'll want to create a system that can capture the insights so
they can be shared with the people who can take action on them.
Visual Data
Navigation in a SmartDoc
Users continue
to demand better user interfaces that make an application easier
to learn, improve productivity, and enable them to gain new insights
and understanding of key data metrics. By improving the interface,
SmartDoc has managed to shorten the "time to enlightenment." The
"Visual User Interface" will enable the user to take a more active
role in the process of visualising and investigating data. The users
interact directly with the on-screen graphics and the data behind
the graphics without having to work with traditional GUI controls.
The aim is to create a data-centric view, in which the user responds
to and interacts with the actual visual presentation of the data
on the screen. The on-screen data object is "live" - the object
itself includes underlying data structures and properties, not merely
a reporting window. The sense of immediacy and speed-of-thought
interaction in both 2D and true 3D space (and correlation between
these two) helps users to gain insight. For example, by clicking
on a graphics object you are able to drill-down to more detailed
information of that object. Other useful interaction techniques
include brushing and hyperlinking. Brushing allows retrieving more
detailed information about parts of visualization without changing
the visualization itself. By moving a pointing device over a particular
component in the visualization extra information appears on top
of the selected object
Multiple Views
Most data are
not best analyzed through the use of a single type of graph. In
order to detect complex patterns within the data it is necessary
to view it through a number of different visualization tools, each
of which is best suited to highlight different patterns and features.
If an information
visualization framework is to handle the broadest range of applications,
it must achieve balance between two- and three-dimensional functionality.
Some problems are best solved with one or the other, however many
require the services of both technologies. We have expended considerable
effort in developing a balanced approach to information visualization.
Where possible, each three-dimensional tool has a two-dimensional
counterpart and the object-oriented nature of the technology ensures
that most functionality is shared between the two. By using the
two together, the power of each is amplified.
The simplest
manner in which to employ one or more graphs simultaneously is to
view them side-by-side. However, the context of each point is lost
in the process. One point may be salient in one graph, but may not
be identifiable in another. Only through interaction may points
like these be located and investigated. The user may select a point
that appears interesting in the line chart to see it labeled in
the scatter plot. The converse may occur as well. This point may
then be deselected from the line chart and the impact upon analysis
performed on the data set as a whole visualized in an entirely different
graph. Therefore, the data, analysis and visualization ' flow' together
in a seamless process of discovery. For example, figure
4 demonstrates correlation between multivariate data visualised
as both 3D spheres and 2D scatter that identifies how they relate
to the each other.
Our components
consistently implement a Model-View-Controller architecture. This
implies a separation between the data and the views and analysis
of those data. This simple and design enables users achieve a high
degree of interactivity with multiple graphs that visualize the
same dynamic data source in multiple 2D or 3D views. The SmartDoc
"Docking Manager UI" component allows you to create multiple views
on multidimensional data sets. Since each of these view shows the
same data, but is otherwise independent form the other views, you
can change the setting for each individual view to highlight a different
aspect of your data. The white area in figure 4
displaying 2D and 3D scatter plots is called the "Visualization
Docking Area." Any opened visualizations will share the available
space in this area. You can re-arrange the position of individual
visualizations by left clicking onto their title bar and drag the
window around. If you have multiple visualizations open at the same
time, you can use this feature to get the layout you desire. This
feature, in combination with dragging visualization windows around
to rearrange them allows you to get nearly any non-overlapping screen
layout you might desire. We have found that this Visualization Docking
Area is a key feature for multidimensional visualization.
 |
| Figure
4: The Docking Manager UI component controls the Visualization
Docking Area. |
Reference Model
for SmartDoc
The basis for
effectively applying all visualization techniques and data handling
capabilities centers on understanding the concepts and operation
of a data flow network (Card et al. 1999), Visual Mappings and the
Scene Tree. We think of the data navigation process as adjustable
mappings from data to visual form to the human perceiver. The diagram
of the mappings in figure 5 serves as a simple
reference model (like other methods). Raw Data, that is, data in
some foreign format, is mapped into ordered Data Tables (Excel spreadsheets).
Data is filtered or clustered. Visual Mappings transform the data
into a Scene Tree. Finally, the Scene Tree is rendered into views
specified by parameters such as camera position, scaling and rotation.
The SmartDoc VUI controls the entire process from data to 3D views.
Our VUI use direct manipulation to interact with "3D Views" (rotation,
zoom, translate), "3D Visuals" (pick any graphics objects, data
correlation) and "Data Cubes" (filter, cluster etc). The data is
stored either remote in the server or locally in the client, while
the visualization process (3D Visual Mapping and Rendering) always
takes place at the client-side.
In the reference
model, visual data navigation takes place at three different stages
in the transformation process. For example, data transformation
can order or classify data differently. By changing the visual mappings
users select different visual forms to view the same data of the
data tables. Finally, altering view transformations include simple
interactions like zooming or modifying the viewpoint in a 3D world
but also user interaction such as brushing.
|
|
| Figure
5: The Visualisation Reference model: The visual data navigation
in a SmartDoc scenario can be described as the mapping of data
to visual form that supports human interaction in a workspace
for3D visual sense making. |
Building a
SmartDoc
In order to
embed an instance of the selected SmartDoc component in, for example,
Microsoft Word, you use the Insert->Object menu. A list of all components
that are installed on your system will be displayed. Select "Scatter3D
Control" and this will create an instance of the control and you
have an interactive visual data navigation application component
embedded in your document.
As Microsoft
Excel is the predominant data source for data in most companies,
we have chosen in our project to use an ExcelInterface component
to import data. This import procedure is done only once; afterwards,
the data is stored as part of the document as a 2D array. The Data
Import wizard will prompt you to open the Excel document, which
contains the multidimensional data you want to visualize. In the
next step you enter the data range that you want to analyse. The
data is normally embedded in the document, but you can also retrieve
the data from a local file or from a central data warehouse.
 |
| Figure
6: The 3D Scatter SmartDoc showing a large multidimensional
molecular data set. Observe the annotated axis displaying names
of the selected molecules. The box to the right is a result
of picking on a sphere. |
SmartDoc will
provide an image of its contents using a bitmap representation.
SmartDoc uses this image for the cached representation of the view
when the component is inactive and for printing the document. Finally,
if the recipient's system does not have the component installed,
the user will at least see the image, though he won't be able to
interact with the data.
When preparing
the SmartDoc, the author can set "bookmarks" or snapshots that highlight
data views of particular interest to different recipients. Colleagues
can use these descriptive bookmarks to quickly locate key information
by simply selecting the SmartDoc view they need. The Snapshot Manager
remembers and records the status of a data navigation experience.
The author has selected suitable data dimensions, display properties,
filtered data with the slide rangers focusing on the data-of-interest
and finally highlighted the "discovery" from a certain angle (viewing
properties) and can now save this status "bookmark" with the Snapshot
Manager. When the document is saved, the Snapshots and data will
be stored as part of the document. When the next user opens the
document, it will start the 3D interactive visualization process
based on the author's bookmarks. The visualization will revert to
exactly the same status as defined by the author. The recipients
can then immediately share the author's insight about data.
SmartViewer
SmartDoc Collaboratories
are sharing an "engine" component responsible for visualisation,
interaction and rendering, called the "SmartViewer". This viewer
component is a central part of the SmartDoc collaborative process
and will become a "freeware" to allow exchange of SmartDoc components
between researchers and engineers. The SmartViewer takes care of
managing all of the visualisation inside the view including taking
full advantage of and manage the complex interactions with the high-performance
graphics layer in OpenGL. The SmartViewer architecture will allow
lightweight SmartDoc components to be deployed across the Web. The
SmartViewer has a size of about 2Mb, while the SmartDoc components
have a small footprint 50-150Kb. Downloading SmartViewer to the
client machine enables this. If SmartViewer is not present, it will
be automatically installed the first time a SmartDoc is downloaded.
All SmartDoc application components share a single installed engine.
SmartDoc in
Practice
We have tested
out a "3D Sphere" SmartDoc on several typical multivariate data
analysis problems. The following example illustrates a typical generic
use - linking between two complementary multivariate data sets as
part of the product design process, one experimental and one theoretical.
Discovering
Bleaching Agents
An important
step in the design of effective laundry detergents is the choice
of bleaching agent. This is the key active ingredient responsible
for the bleaching effect. Bleaching is a molecular not a physical
process, the beach chemically interacts with the stain or fabric
surface releasing the stain species.
 |
| Figure
7: Bleaching Processes |
The overall
effectiveness of a possible active will therefore depend on three
things:
- Its molecular
descriptors (electronic, electrostatic, structural etc).
- The in use
conditions for example the active concentration, the pH, presence
of other active species, or the length of time the fabric is exposed
to the bleach.
- Ultimately
the effectiveness is judged by the consumer who visually assesses
the appearance (and possibly other parameters).
- Our problem
then is to link data describing these three areas with the aim
of perhaps identifying novel new actives or optimum conditions
for effectiveness for existing ones.
In the example
here, a large database of molecules is screened for effectiveness
in removing dissolving tomato oil stains. This is an initial screen;
the stains are simply fixed in solution they are not attached to
a fabric (any molecule showing some effectiveness in this first
screen would be tested on actual fabric stains in a similar way)
The molecular
descriptors for each candidate are calculated using quantum mechanical
techniques, these descriptors fall into two classes being either
about the electronic properties of the molecule (e.g. the chemical
bond properties) or its structural properties (its shape), both
are likely to matter in bleaching. We note in passing that for a
large number of candidate molecules, this data set alone is very
large and is ideally suited for exploration using the navigator,
should the user be interested finding correlations among molecular
stuctures alone.
 |
| Figure
8: The Data Discriminator "Filter" and molecular descriptors.
The filter component is used to reduce the data displayed in
the SmartDoc. The filter menu will open a window with range
sliders, one for each column in your data file. This allows
you to adjust the minimum and maximum data values which are
displayed for each column. A data value (a glyph representing
a row in the spreadsheet) will disappear from your visualization
if any one of the values in this row is blanked out by the slider
associated to it. The data contains electronic and structural
descriptors of a set of candidate bleach agent molecules |
Collecting data
on the in use conditions and the consumer perception is classically
much more time consuming. To speed things up a high throughput methodology
is adopted using a 96 well plate. This is an array of small cells
each of which can contain a few millilitres of a TOL sample at a
given concentration pH etc. At time = 0 a candidate bleaching molecule
is introduced into each cell and any colour change monitored as
a function of time. The bigger the colour change the more effective
the molecule as a bleaching agent for the given stain, this is the
measure we use as the substitute for consumer perception. Such an
experiment can be highly automated and repeated continuously.
 |
| Figure
9: High throughput screening. A time series signal is recorded
from each cell in the 96 well plate. In this case the cells
are lit from below and the signal is proportional to the amount
of light transmitted through the sample, so the higher the signal
the more effective is the candidate bleach agent in dissolving
the stain. |
The data is
collected within an Excel spreadsheet and merged with the molecular
properties data. We then have a multivariate time series data set
containing molecular, in use and bleach effectiveness information
to investigate.
Loading the
data into the navigator using the import wizard gives us an initial
view, which, for a 3D scatterplot uses the first three columns of
data to establish the axes system. The first three columns in this
case are appearance measures under three different in use conditions,
drilldown allows us to find out every property associated with the
contents of a particular cell in the 64 well plate
 |
| Figure
10: Initial data import, each glyph relates to the contents
of one particular cell |
What we are
initially viewing is part of the space of "in use "conditions. Our
first aim is to try and identify the type of molecule that might
be a good detergent active so we first change interactively to a
view where the axes are defined using three of the molecular descriptors.
Any point within the axes space then represents a particular molecular
configuration existing or not. We can then map the bleaching effectiveness
under a given set of conditions onto this space of molecules and
instantly see the performance of each active molecule. Mapping on
different bleach times, concentrations and other experimental conditions
enables us to quickly confirm the general rules for bleaching effectiveness,
for example:
Longer time
-> better bleaching
Higher concentration
-> better bleaching
The three molecular
properties we happened to choose here to define the axes clearly
show no correlation, however we can map a further molecular property
into the visualisation by using the glyph size variable. The application
performance enables us to quickly scan through all the calculated
molecular properties, something interesting happens when we map
on the ability of the molecule to act as a accept or donate electrons
from/to its environment to the glyph size
 |
| Figure
11: After swapping the axes so that they are now represent three
of the molecular descriptors we can use the glyph colour to
represent the amount of stain removed after a certain in use
condition (in this case the amount of stain removed after 1
week at pH 10 without any additional accelerator molecule).
In these conditions the successful candidate molecules are highlighted
yellow-red. There is no real clustering either of colour or
position indicating that the molecular descriptors used for
the axes do not correlate well with the molecular bleaching
capability. |
 |
| Figure
12. Clearly the vast bulk of effective bleaching agents have
similar -rather limited electron donator/acceptor properties.
|
 |
| Figure
13: Exploring the molecular descriptor space further allows
us to pin down the successful candidate molecules to a much
more limited region of the configuration space. Colour here
again is a measure of the effectiveness of the molecule at bleach
removal |
Clearly for
most cases a strong correlation between these electron acceptance/donation
properties and the effectiveness of bleaching. This leads us to
attempt to produce a more effective visual clustering by making
the electron acceptor property one of the spatial axes. Further
sampling of the molecular descriptors enables us to cluster the
effective candidate molecules into a much smaller region of the
configuration space.
 |
| Figure
14. Two coupled scatterplots, which illustrate the effect of
the additional additive. In both the glyphs are coloured by
the signal strength with additives, the first plot (left) is
of the space of signal strengths without additives, if the additive
made no significant difference we would expect the colours smoothly
vary from blue at origin to red at the extremes-but there are
clearly two clusters. Further, by picking we can see how these
clusters map into the molecular descriptor space.
|
A SMARTDOC
SCENARIO
Step1: Create
the Spreadsheet data. Transform raw data into data tables.
Step2: Select
and download visual data navigator application components "Collaboratories"
from a central portal component repository. Download the SmartViewer
if not already installed.
Step3: Embed
one or several selected Collaboratories in, for example, a Word
document by using the Insert->Object menu. A list of all components
that are installed on your system will be displayed. Select "Scatter3D
Control" and this will create an instance of the control and you
have an interactive visual data navigation application component
embedded in your document.
Step4:
Open the Collaboratory and visualize interactivily the data inside
the document. View, zoom, select attributes, pick data and isolate
data within certain ranges by filters until the author has discovered
pattern or trend in his data. Save snapshots at each level of discovery.
Step5:
The author is now ready to create the final SmartDoc document. Text
(content), Final Image (selected view of data), Statusfile for the
snapshot (selected data and visualization attributes are saved),
Visualization components (pointers), Data (embedded as an "array").
The author sets snapshots that highlight data views of particular
interest to different recipients. Colleagues and clients can use
these descriptive bookmarks to quickly locate key information by
simply selecting the report view they need. Recipients also have
the flexibility to create their own bookmarks for at-a-glance reference
and comparison.
Step6:
After the author completes his report, he sends it to both internal
and external constituents who have access to the Viewer. He posts
the document to the organization's Web site. Now, internal managers
and marketing researchers etc can interact with the document on
the organization's Intranet while external clients can work with
the document via Internet.
Step7:
Recipients retrieve the SmartDoc document from the Web site. The
visualization components and Viewer are downloaded to the client
machine (if not already available).
Step8:
Recipients can now dig into results and get more value by creating
new meaning and understanding in the results.
- Start the
analysis from a snapshot
- View the
analysis behind the report
- Interact
with the results and digest
- Change visualisation
parameters
- Discover
new meanings of data, trends and correlation
- Recipients
are more active than with just text reports
- Promote
collaborate research
- Generate
a new SmartDoc report
CONCLUSION
The prospect
for SmartDoc and Collaboratories look promising. In particular,
the Web has created a model for disseminating information. The traditional
"thin" client-server model, however, works poorly for data interactivity
required in a data navigation scenario. The standard Web interface
encourages data navigation and its visual user interfaces to focus
on lowest common denominator, which would require high bandwidth.
The emerging visualization component architecture based on "atomic"
components with a small footprint could, however, move data navigation
techniques from research to products.
The SmartDoc
EC funded research project has demonstrated how to deliver an interactive
experience in an electronic document on the Web based on embedded
interactive visual data navigation components "Collaboratories".
Upon receipt of a SmartDoc, the recipients can immediately share
the author's insight and zoom-in on specific results that affect
their activities. Clients and colleagues can rearrange multivariate
data to uncover patterns and trends that may go unnoticed in lengthy
paper reports. The distributed architecture is based on "Application
Component Sharing" providing real-time data interactivity, reducing
the load on the network and with zero-administration client deployment.
Also important
is the aim to promote the use of a component-based approach to the
development and engineering of software systems, applications and
services (figure 2). Customisable and scalable
high-level "application" and "functional" components were designed
and developed from low-level "atomic" components. Our Collaboratories
are based on Advanced Visual Systems' OpenViz, a low-level visualisation
component framework. Atomic components from several other sources,
including data interactors, data filters, analysis, and data access
were also integrated. We believe that using lower-level atomic components
for developing application components would provide better scalability
and more customisable visual data navigator components. Atomic COTS
components from different vendors (or developed when necessary)
were used in assembling functional and application components.
We have shown
the possibility of deploying 3D visualisation in electronic documents.
Based on our experience, we have drawn some tentative conclusions
regarding 2D versus 3D data visualisation. We can conclude that
2D data visualisation methods are more easily accessible to the
user. The 3D data visualisation allow the user to combine more information
into a single scene, but these methods are not yet accepted as instruments
for decision making among the business community.
Another overall
goal of SmartDoc is to make people more effective in their information
or communication tasks by reducing learning times, speeding performance,
lowering error rates, facilitating retention and increasing subjective
satisfaction. We believe that customisable and scalable Visual User
Interface (VUI) components in collaborative work can increase effectiveness
for users who range from novices to experts and who are in diverse
cultures with varying educational backgrounds.
SmartDoc will
be tested and validated in European global industries with geographically
distributed remote users, linking people and their desktops (Collaboratories)
in a worldwide "Virtual Data Environment". Customised medical imaging
application Collaboratories will also be tested in European healthcare.
We believe that SmartDoc can contribute to a new wave of accepting
advanced real-time data visualisation and navigation in a collaborative
environment with the focus being on the people using the system.
SmartDoc's features
include:
- Dynamic
multi-dimensional graphics that users can interact with in real
time
- Rich set
of component resources with granular control of details
- Platform/rendering
library independence through SmartViewer Plug-in - Visual User
Interface
- Tight integration
between data and visualisation objects
- Optimised
rendering based on scene tree hierarchical graphics structure
- Binning,
filter, crop, sort, and aggregation to achieve good interactive
response
- Drag, rotate,
zoom, pick empowers users to explore data
- Data reduction
integrated into visual data navigation
- Full integration
with the data warehouse or spreadsheet
- Presentation
graphics with axes, legends, annotation and high-resolution hard
copy
Acknowledgments
This paper was
supported by the European Community in the ESPRIT Project CONTENTS
(EP 29732) and Advanced Visual Systems.
|
