Monday, March 17, 2008

Visualize Muckety



I have seen a lot of research on web visualization and I have seen a couple of companies spring up whose product was based around using various visualization techniques to provide unique perspectives or insight. These products or services can provide decision makers with specialized tools that help interpret information and understand relationships.

Combined with Internet technologies, many displays approach the idea of overview by given dashboard configurations of symbols or distillations of large quantities of information into more easily digested representations or concepts.

When they are interactive (as they should be), they allow you to determine the path for that exploration.

I have talked about tag clouds on this blog before, and have even included recently a tag cloud widget in the sidebar. The widget does not give complete coverage of the tags I use here, but it provides a second visual example of topics that draw from a sampling of the categories I’ve created.

Recently that I have seen more usage of these topical visualization tools on the web and in particular regarding news stories. I wanted to use CNET’s use of Radial Hierarchical Networks methods for there visualization of top stories in “The Big Picture” as a live example, but unfortunately I can’t find it anymore.

However, for a fun exercise in interactive representations of how the news connects, check out www.muckety.com. It uses the same approach that CNET’s Big Picture did and reminds me of the old Persuadio maps.

Here is what Muckety has to say about itself:

Muckety is published by Muckety LLC, a company founded in 2006 by a team with years of experience in journalism, technology and online publishing.
The name Muckety derives, of course, from muckety mucks. Some follow the money. We follow the muckety, producing a daily news and information site based on online databases (which we enlarge daily), extensive research and old-fashioned journalism.

The founders/editors of the site are Laurie Bennett, Gary Jacobson and John Decker, who are all award winning journalists with expertise in graphics, photography and reporting.

Using the web site and control clicking on Laurie’s name I get a visual of her relationship to a multitude of new organizations where she acted as a reporter, editor or co-founder in one case. In the same graphic, it also provided relationship information to family as well as work.

Sadly, I found that I had no relationship to anything when I searched on my name. But as the site is updated and expanded on a continual basis, one day I may relate to something.

For a quick explanation of how to use it, www.visualcomplexity.com, an amazing site which is a great stopping place for information and examples in graphical representation explains.

After performing a search and choosing from the search results, you'll get a map that shows connections between people and organizations. To view descriptions of the relations, click anywhere on the map to activate it. When you pass your mouse over connecting lines, you'll see a popup box describing the connections. Solid lines represent current relationships; dotted lines show former relationships. You can also re-center a map around one or more players by clicking on the map background and dragging your mouse to draw a temporary box around them. The chosen boxes will be highlighted in pink.



For a puzzle to solve, use Muckety to connect how Groucho Marx and Frank Sinatra are related.

4 comments:

Andrew Fry said...

If it was too difficult (or you are too lazy) to find the connection, do the following.

Double click on Zeppo Marx, then double click on Barbara Blakely.

There you go.

Jeff Becker said...

I think this has a lot of potential, but it has a long way to go.

One problem is the heuristics of compiling a large number of various pieces of data into one representative subset. Imagine trying to compile all of the data relating to, say, "the president of the United States". Start with 43 presidents then include their vice-presidents, cabinets, congressional leaders that they worked with (and against), foreign leaders on the scene at the time, opponents in the campaigns, etc. and the interconnections of all of those individuals, each being weighted according to their proximity and importance. Throw in the political history of each president for each office they held leading up to their election as president. Then doing the same for the other person you're trying to connect them with, but roughly in reverse, so that the interconnections can be found. And how many degrees of separation you're looking for makes a big difference. Doing this for even one president might take a moment or two even on a fast computer. With that problem comes the shortcomings of currently available database platforms which are relatively linear in their ability to interrelate data.

Another issue is the visual representation of numerous interconnected pieces of data and their relations in an easy to understand format. Just connecting one president to one congressman of the same party might fill up your screen with an indistinguishable mass of icons and text.

As an example, type in your favorite recent president and click on a connected 'muckety muck' and continue on with that for even two or three levels.

I'm sure someone will come up with something soon and create the next Google.

I'm currently working on a much smaller process for an open source project. It can be somewhat mind-bending at times. But, I think the complexity makes it that much more fun.

Andrew Fry said...

I'd like to know more about your open source project. If you have a moment contact me with the information.

As for the effectiveness of this type of visualization approach, it definitely will increase in value as technology improves. Filtering and semantic associations will be an incredible step forward. The idea of moving through a three dimensional space of association will change the whole ball game.

With the right semantic engine and filters, you will be able to adjust and control the amount of depth and breadth of related information to suit your taste.

Jeff Becker said...

I'll e-mail you about the project.

Filtering and semantic engines, as the processes develop and improve, will certainly improve both the quality and quantity of the resulting associations.

However, each stage and application of filtering and (semantic) analysis introduces it's own level of error. And as semantic analysis rises above simple grammatical analysis, requiring more artificial intelligence, it also increasingly introduces levels of subjectivity.

Then, once all of this has taken place, selecting the relationships between the subjects of the data analyzed for storage and reference, as well as the relationships between that and the query introduced, requires not only highly complex algorithms that must apply subjective weights, it must also analyze the subjectivity of the query and compare it against the subjectivity of the data being analyzed for inclusion in the result.

As an example, picture a well educated political analyst looking for a connection between an opposing candidate and a politician involved in a breaking scandal. He'll want to ignore the rantings of moonbats (on either side) and focus on information that produces actual ties between the two. However, the actual tie might be hidden in an emotional rant that the system may have disregarded because it evaluated the emotional language as the ranting of a moonbat.

And from the beginning, all of this is based on the grammar of the sentences introduced into the processes.

In short, filtering and semantic engines will get the concept over the first hurdle on it's way to usefulness.

Obviously, this are just a few of many areas of development that I've been keeping an eye on over the years. I think the most interesting part is seeing which sector advancements come from. Will they come from garages and basements, research schools or corporate labs?