AW: [topicmapmail] Plumb Visual Thesaurus Redesigned

Rath, Holger (empolis DA) Holger.Rath@empolis.com
Tue, 29 Oct 2002 13:09:01 +0100


Hi,

Steve Pepper wrote:
> At 10:10 29.10.2002 +0100, Rath, Holger (empolis DA) wrote:
> >Concerning the graphical visualization of Plumb
> >Design, I have to say that it is - for sure - an
> >eye catcher. But it is not really easy and efficient
> >to use. I prefer the StarTree from inxight much more.
> >It does not look that sexy but offers many advantages.
> >BTW: That's the reason why we integrated it into k42 -
> >our Topic Map engine.
> 
> It would be interesting to know what you believe those
> advantages are. 

Before I list the advantages let me write about the different
categories for TM rendition and visualization:

1- HTML
   a- general TM browser like k42 and OmniGator
   b- web applications built on top of TM engines adressing 
      the specific needs of the application domain and its users

2- Visualization of *complete* TM

   Purpose is to get an overview how the TM looks.
   Clustering of "similar" topics is a big issue as well
   as an visualization algorithm that scales up to many
   many topics and assocs (millions!).

   There are some research results available for these kinds
   of visualizations. I am not aware of a product.

3- Visualization of a "topic in its context"

   Purpose is to get informed about the information related
   to a certain topic by following the multi-dimensional
   links coded in the TM.

   This is called "incremental explorative browsing" of ontologies.
   Something which is seen as an important function by certain
   analyst groups. Explorative browsing is - as the name says -
   interactive exploration of the coded knowledge. It is not very
   useful if you are looking for something specific - you should
   better perform a search to get to the required information. But
   it is very good in when you want to "learn" something from the
   TM or just be "inspired" by the TM - without knowing in advance
   what you are really looking for. It is nearly like an advanture 
   game.

StarTree and others visualization tool like TheBrain and ThinkMap are
fitting in the 3rd category.

The advantages of StarTree against TheBrain and ThinkMap are
(in my opinion):

- Users can utilize their visual memory to remember where a certain
topic is (as long as you start with the same root topic and the map
has not changed, of course). Other tools contineously change the layout
of the graph which makes it impossible (!) to use our very good
visual memory.

- Because the TM graph is drawn as a tree you don't have crossing lines.
The disadvantage that an arbitrary graph is "reduced" to a tree is
there, but compared to the other advantages acceptable. All loops in
the graph are rendered as copied subtrees, which means that no information
is lost - the users always sees everything, which is related
to the topic. BTW: inxight promised to improve the StarTree by a feature
to draw lines on-the-fly connecting dublicated topics to inform the user
that he is looking at something also visible somewhere else. And: many
TMs will for example model taxonomies which are - mostly - strictly
hierarchical.

- The StarTree only moves when user drags the mouse.

- StarTree is able to browser to the map without re-focussing
and redrawing the graph. This means that you can navigate through
your TM in one "direction of interest" and come immediately back to the
starting point just with a simple and quick movement of the mouse.
This features makes StarTree extremely scaleable to draw thousands
of topics and assocs at the same time.

- User always know where he is coming from, because you always see where
you started.

> So far I haven't been very impressed by
> StarTree (or any other graphic visualization of topic
> maps).

I was skeptical for a long time as well - you probably remember
that, Steve. But most of our customers really want to "see"
the TM graph and navigate it. That is the big advantage over
many other KM technologies using implicit knowledge models,
which cannot be visualized (e.g., statistical approaches).
It is true "A picture says more than thousands words".

> I wouldn't say that StarTree is any easier or more efficient
> to use than ThinkMap: 

Might be true that I know how to use StarTree because I use
it very often

> They both satisfy the urge for what
> Erik Naggum used to call "instant visual gratification" ...
> and they both drive users bonkers after a very short time.
> (One person who used StarTree told me she felt "intense
> seasickness" after just a few minutes.)

Well, what drives the users bonkers? The kind of visualization or the
information rendered by the visualization? I think it is a combination of
both. Just navigating an HTML-rendered TM is also quite boring after a while
- everything just looks so similar and is completely redundance-free. But
humans are used to redundancy (spoken and written language has lots of
redundancy) and run into uncomfortable feeling when it is missing.
This could have been also a reason for the "seasickness".

> I think more work needs to be done on algorithms for
> reducing the complexity of a topic map before it makes
> sense to try and visualize it graphically, otherwise the
> result is unusable.

Every graphical visualization and not only of TMs but in general
is easier to grasp by humans than text as long as it does not get
to complicated. I agree that a topic with more than 50 assocs just does not
look good in StarTree. This is where clustering an similar approaches
enter the scene. And I agree that more research has to be done
to find proper clustering algorithms. Might be that they are
quite application specific.

Regards,
--Holger