Wednesday, May 27, 2009

Visual Analysis Everywhere! (Original Research And NVAC.pnl.gov)

I have finally gotten over enough of my jet lag to get back to work and it seems like everywhere I turn there is something interesting cropping up on visual analysis.

Two projects, in particular, recently crossed my desk, one by a student in my Advanced Analytic Techniques class, Andrew Canfield, and the other from the US's premier resource on visual analysis, the National Visualization and Analytics Center at Pacific Northwest National Laboratories.

Visual analysis allows people to more easily identify patterns in the massive data flows that are pervasive today. There are a wide variety of tools available for this purpose but Andrew focused on one, Starlight, to conduct his preliminary analysis (which he conveniently documented in the YouTube video below) of Turkish foreign policy with respect to the Caucasus.



While subject matter experts on Turkey might disagree with Andrew's conclusions, the real intent of his efforts was to experiment with the method. I think his video gives a fine introduction to some of the benefits of visual analysis.

In addition to Andrew's work, the NVAC also recently published their quarterly magazine on visual analytics, called VAC Views. Under the heading of full disclosure, I suppose I have to mention that Mercyhurst gets a nice write-up (on page 19 for those of you keeping score at home...). Beyond the shameless self-promotion, however, there are a number of articles that are quite interesting, not the least of which is NVAC's taxonomy of visual analytics. Previous issues of VAC Views contain other, similar, gems.

The rest of the NVAC site is also worth checking out. It contains a wide variety of videos and other resources including the invaluable Illuminating The Path e-book. Much more than the research agenda it describes itself as, Illuminating The Path is an essential background reference for anyone interested in visual analysis.

Reblog this post [with Zemanta]

1 comment:

Pat said...

It seems like an interesting way to use clusters and categories a la Black Swan, with the additional value of the more fundamental content analysis. Kudos to Andrew.

In my opinion, users of this tool should avoid using too much data collected from news sources. After many years of reading the NYTimes on the train, I can tell you that news articles almost always offer a small piece of news framed within a relatively biased context section that helps the reader put the story in perspective. That context section can be repeated over and over again, article after article, distorting the results of visual analysis by the inclusion of terms unrelated to the news story. The media add context because they think the reader needs to know what they know to understand a bit of news. It's like Turkey 101, every time they write a piece on Turkey.

Maybe data from pdf documents would provide cleaner data? If news feeds are a necessary evil, I would suggest tagging the news portion of each article for the data base and ignoring the context sections whenever possible.

Again, I'm impressed with the use of this analytic tool and I'll seek to apply it myself.