My Photo

Your email address:

Powered by FeedBlitz

April 2018

Sun Mon Tue Wed Thu Fri Sat
1 2 3 4 5 6 7
8 9 10 11 12 13 14
15 16 17 18 19 20 21
22 23 24 25 26 27 28
29 30          
Blog powered by Typepad

Become a Fan

« The Dissenting Numerati | Main | Santa's Surveillance Operations Center Enjoys Big Gains in 2008(*) »

November 24, 2008


Feed You can follow this conversation by subscribing to the comment feed for this post.

Jay Levitt

I think the hardest part of explaining this is the cliche problem:

Puzzles are, in fact, a perfect analogy to contextualization. But we've been talking about "another piece of the puzzle" for decades, or maybe centuries. When you say "it's like putting together a puzzle", people no longer get a mental picture of puzzle pieces; it's not a metaphor at all. It's just an idiom, like "momentum" or "perspective" or "focus".

Once it becomes idiom, you lose all the cool facets that came with the metaphor. Puzzle pieces are part of a whole! And if you look them from a different angle, you might mistake them for something else! And they only fit together a certain way! And... doesn't matter. The metaphor has lost its power - quite *because* it is so very apt.

This is why you don't want logos or stock art involving globes, rainbows, multi-racial handshakes, and smiling people on headsets. It's not that they don't fit; it's that they don't communicate anything anymore.

Finding a good metaphor is no longer enough. You have to find a fresh one, too. That's much harder - in fact, it's like finding a needle in a haystack.

Baris Yazici

Is there are real-life problem are you trying to generate solutions for?

I am asking, because, sometimes, opposite of context accumulation (context decumulation?!??!) may also make sense. What I mean is, instead of starting with zero context and try to build it up, you can start with maximum possible context and try to reduce it to only one (the exact location and position of the puzzle piece relative to others).

Induction vs. deduction.

Of course, this is only useful when you know all possible locations and positions your puzzle piece can be initially and from that remove those that it can't be in.

Sounds like a brute force algorithm, but it is not.

Actually, this is exactly the algorithm I had used in the new Address Parser in EAS to make sense of human language addresses (IBM patent pending).

Brand Hunt

Good metaphor! I love visualizing multiple puzzles in one box with no pictures -- seems like there may be a market there? :-)

IMO, for the puzzle analogy to be similar in difficulty you'd have to have a tessellation puzzle where there'd be no physical limitations to putting pieces together (i.e. no hammer and no weighting towards false negatives). I often see two puzzle pieces that look like they should be adjacent, only later to find that I'd need the hammer -- I think the puzzles have more "context per piece" than typically exists in your systems.

Jeff Carr

Terrific post, and comments. You've inspired me to re-think some sticky issues by using Context in a different way - recognizing the absence of something that should be present, but isn't.


I came across your blog after googling information management. This seems to be used extensively in the insurance industry:

If the detailed explanation in the above site is any indication, the average Joe doesn't have a chance.

How does one get to the heart of these processes?

Randy Brokaw

Good stuff. Thought provoking.
Your observation about iteratively making assertions, match, no match, or possible match, is similar to the InfoBright technology for BI and the underlying rough set concepts.
It is interesting to note that the premise of their BI architecture, incremental, scalable updates to a warehouse without reloads, is similar if not identical to the theme of your post. It seems the hurdles of existing warehouses have influentially formed the zeitgeist.



I know it's only an analogy, and not meant to be precise in every detail, and I'm also not trying to argue against your underlying convictions - which I think I agree with.

But another way to look at the end game of putting a puzzle together is to regard pieces that are already joined as being a single (composite) piece. Looked at from that POV, the process of assembling the puzzle has monotonically decreasing difficulty - all the way to the point where you're down to the last two pieces (possibly both composite), at which point a 2yr old child can finish the puzzle.

In other words, the implied(?) argument that things might be expected to be getting harder because you have more pieces on the table, can be flipped on its head - with great effect, because the puzzle *is* getting easier, as seen by the speedup provided by non-false-positive context.

In a sense I'm just picking nits. But I do also find great value in thinking about practical / real-world analogies and how they can give insight into how we go about designing computational systems. FluidDB is in large part about always being able to insert new information into a system in the place(s) where it (is imagined it) will be of most value. It's about how information becomes more valuable when it's in the right context. I love thinking about how we work with information in the real world (e.g., putting a puzzle together, or even just putting a bookmark in a book or a post-it note onto something) and comparing it to the way our familiar computational environments force us to work with information. The gap seems vast, as I think you agree. I also find it fun to think about how we might build systems that make the latter more like the former, and to have also dedicated so much of my life to actually trying to build such things.

Cheers from NYC.

The comments to this entry are closed.