But so-called visual analytics will inevitably work its way down to ordinary business and administrative IT users, he says.
Thomas is chief scientist for information technologies at Battelle Pacific Northwest National Laboratory in the US, one of nine US Department of Energy "multi-programme" national laboratories, and one of six consulting scientists for the US Department of Homeland Security. As a field with strong relevance to human behaviour and cognition, visual analytics research benefits from an infusion of different cultural views from around the world, he says.
One of the most pressing tasks in visual analytics is to set up a research agenda. Here the input of New Zealand institutions would be welcome, says Thomas.
While in New Zealand last week looking at how US and New Zealand experts might co-operate, Thomas visited the Human Interface Technology Laboratory (HITlab) in Christchurch and Wellington’s Media Lab South Pacific.
Representing data in graphical form is a discipline some 40 years old, but what it lacks is the element of rapid interaction over time, Thomas says. Visual analytics seeks to set up a “highly interactive discourse” between the worker and the data, the worker framing hypotheses based on the visible data and it being analysed and presented quickly another way to test and refine these hypotheses.
Data mining might be the first business application that one would imagine as an application, but Thomas suggests the media. A journalist seeking to frame a news story would gather all information on the matter and any previous stories written about it, then put together a visual presentation that highlights the important elements of the story and the way they fit together. Just as important, it would show where the gaps are in existing information and previous coverage, so as to get a “new angle” on the matter.
Typically, a visual analytics application will deal with very large volumes of data, examined in multiple dimensions; “200 is about the norm”, Thomas says, adding that the human mind seems surprisingly capable of dealing with information of that complexity.
Where a simple retrieval exercise from a business IT system or an internet search engine would present only the findings on one topic, the kind of tools his laboratory is developing enable results to be retrieved on a number of topics. Overlapping subjects that they all mention can be identified for further exploration.
An already-commercialised product from the lab’s work called In-Spire maps such “galaxies” of word and concept coincidence in texts. It is now being applied to DNA sequences.
Such a system must be capable of dealing both with structured data such as customer records and unstructured data; for example, emails.
The concept of a Theme River is also being studied; a way to represent the rise and fall of themes in a sequence of documents or speeches over weeks or months, to see the way an administration’s policy is developing, or perhaps patterns of concern among the customers of a business.
One of Thomas’s priorities for visual analytics platforms is that they be “walk-up usable” -- easy to use with natural movements of hand and head and without mechanical pointing devices which “tether” the operator.
Graphical presentation may be better done on a horizontal table rather than a vertical screen, particularly when a team is working on the analysis together, he says. “Immersive” environments which surround the operator in a “cave” of visual presentation screens have also been found to be productive.