Quick PhD description
The field of Information Visualisation (IV) is concerned with reducing the cognitive overheads of understanding complex information structures through the use of visual representations. IV achieves this by utilising appropriate visual metaphors composed of representative and interactive elements that take advantage of the strengths of human visual perception.
IV differs from Scientific Visualisation (SV) in that it deals with mainly non-physical data sets and is applied to more abstract structures such as networks, derived high-dimensional data, and 'tree' structures.
Tree structures have been the focus of many IV efforts such as Robertson et al's Cone Trees, and Shneiderman and Johnson's TreeMaps (see downloadable refs at bottom of page). However, most of these efforts concentrate on visualising the structure and properties of just a single tree. The primary aim of my PhD is to develop a metaphor that will allow users to explore the relations and properties of and across multiple tree structures.
Currently, we are using data sets taken from the field of botanical taxonomy supplied by taxonomists at the Royal Botanical Garden Edinburgh (RBGE). I will be designing and constructing visualisations of successive generations of botanical taxonomies, with particular emphasis on the structural modifications that occur between them. It should enable the taxonomists to follow the progress of a species, or a genus, or a family etc, and the general ideas should be applicable to other domains too.
More info on my Phd can be gained by downloading my 1st year report.
Initially, we came up with two competing visualisation metaphors, one a graph-based visualisation, and the other a set-based visualisation, details on which can be examined in my 1999 June summary and July summary.
We performed user tests on the graph-based and set-based visualisations, and came to the conclusion that the graph-based visualisation is no longer worth pursuing due to a number of factors. These included slow redraw speeds which affected user interaction, and that even with basic filtering it tended to overwhelm the user with information.
A second raft of user tests performed on the remaining visualisation (only set-based vis left now), after all the perceived necessary changes from the first user test were introduced, and more and larger data sets were included in the visualisation (7 hierarchies instead of 4). This test was to ensure that the functionality required by the taxonomists was in place i.e. could they do all the tasks they had asked to be supported, even if the usability may have been a bit unforgiving.
Necessary changes to the visualisation were identified from the second user test and these were included in the updated visualisation. Screenshots from various stages of development can be viewed below.
I attended the AVI 2000 conference in Palermo, where I presented a paper and gave a demonstration of the prototypes to interested parties.
We began to think about how to cope with larger data sets, with regard to running out of screen space. A focus and context technique will need to be developed to meet the demands of such data sets.
In July 2000, the third set of user tests were performed at the RBGE, using the applet version of the prototype (so I didn't have to spend time setting up software over there). These were performed with the previous subjects, and with some new users, and was mainly to test for usability issues i.e. could they see how to do the supported tasks? And could they interpret the visualisation correctly?
Kent Wittenburg has supplied us with an alternative data set to test out the limitations of our visualisation prototype, namely a set of multi-dimensional product data for digital cameras.
ScreenshotsClick on a thumbnail picture for a larger image. Each thumbnail is about a 20% by 20% proportion of the larger image.
|Early graph - unfiltered (42 KB)||At User Test 1 (Sept 99) (34 KB)|
|With control panel (39 KB)||Last Graph Version
An Overview of the 800-node graph (64 KB)
|Last Graph Version
Examination of Drude's 'Unplaced' category (58 KB)
|Last Graph Version
Examination of Caucalis (60 KB)
|Last Graph Version
Globba Revision data set (100 KB)
|Early Sets - 1 Selection (18 KB)||At 1st user test
Sibling distribution for Molopospermum (36 KB)
March 2000 (68 KB)
|Set Visualisation - 8th Sept 2000 (37 KB)
DOI of a group set to max by any leaf being selected in a group
|Comparison by Tribe rank
after removal of lower categories (51 KB)
(Supplied by Kent Wittenburg) (62 KB)
|Set Visualisation - 25th Sept 2000 (61 KB)
DOI of a group decided by average of individual leaf DOIs
|Set Visualisation - 30th Oct 2000 (59 KB)
DOI of non-selected members of a group decided by ratio of member nodes selected divided by total nodes selected in last selection. DOI of group then takes average of all member nodes (selected and unselected).
|Set Visualisation - 13th Feb 2001 (45 KB)
Globba revision data set. New display algorithm allocates space to leaf groups according to relative DOI totals.
|Set Visualisation - 15th Nov 2001 (51 KB)
Apiaceae data set with quantised space allocation (see IV2001 paper ref below). Fairer to small groups of nodes and more even in sizing selected nodes.
|Set Visualisation - 22nd Nov 2001 (90 KB)
Apiaceae data set. Montage of effect of rank hiding in visualisation.
Apiaceae - Multiple Trees
Apiaceae - Graph
Globba Data Set - Multiple Trees
Globba Data Set - Graph
Brief instructions -- Link Fix -- (348 KB) on the applet.
My guide to to tracking down academic papers
|AVI 2000 Presentation from martinjgraham||UIDIS '99 Presentation from martinjgraham|