[seek-kr-sms] RE: growl+icons

[Joseph Goguen]

Hi, ive got to run off to our dept retreat, but wanted to add just a few
random thoughts to this discussion, since i teach the user interface design
courses and we talk a lot about icons and graphs and information visualization
in general (indeed, ive written 3 or 4 papers on IV, and with Kai Lin, worked
on a system that displays proofs).

Shawn is very right that what works well depends crucially on the user, and so
we should be looking at what actual users will want to do with the system;
computer scientists often have quite different intuitions about interfaces
than domain scientists (see extensive discussions in Shneiderman's book, for
example, or almost any other).  Direct manipulation interfaces, based on
visualizations that preserve as much pertinent structure as possible, and
display it in an intuitive way (for a given class of users!) is the best way
to go; just cause something is "visual" doesnt mean its necessarily better.

Large ontologies logical structures tend towards heterogeneity, and so it can
be very difficult to know where you are in them, and hence difficult to
support direct manipulation; Kai and i found this in our Kumo proof system.
Some kind of browser technology is appealing in the case, taking advantage of
the modularity of logic.

For OWL, definitely class hierarchies are a good idea, but one doesnt need the
entire lattice, just the Hasse diagram of the given implications would be
enough, and would reduce the clutter for large ontologies.  It is probably not
a good idea to try to reduce logical formulas to graphical systems of icons.

If we had time and money, it would be good to hire a cog sci student to work
on this; i wish i could do it myself, it is a lot of fun, but dont have the
time, though i would be glad to offer advice, literature references, etc.

Cheers,

  joseph



>Delivered-To: seek-kr-sms at ecoinformatics.org
>From: Shawn Bowers <bowers at sdsc.edu>
>User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.5) Gecko/20031007
>X-Accept-Language: en-us, en
>Cc: seek-kr-sms at ecoinformatics.org, "'Kai Lin'" <klin at sdsc.edu>
>X-BeenThere: seek-kr-sms at ecoinformatics.org
>X-Mailman-Version: 2.0.13
>Precedence: bulk
>List-Help: <mailto:seek-kr-sms-request at ecoinformatics.org?subject=help>
>List-Post: <mailto:seek-kr-sms at ecoinformatics.org>
>List-Subscribe: <http://www.ecoinformatics.org/mailman/listinfo/seek-kr-sms>,
>	<mailto:seek-kr-sms-request at ecoinformatics.org?subject=subscribe>
>List-Id: <seek-kr-sms.ecoinformatics.org>
>List-Unsubscribe: <http://www.ecoinformatics.org/mailman/listinfo/seek-kr-sms>,
>	<mailto:seek-kr-sms-request at ecoinformatics.org?subject=unsubscribe>
>List-Archive: <http://www.ecoinformatics.org/pipermail/seek-kr-sms/>
>Date: Wed, 07 Apr 2004 13:03:15 -0700
>X-Spam-Flag: Spam NO
>X-Scanned-By: milter-spamc/0.15.245 (fast.ucsd.edu [132.239.15.4]); pass=YES; Wed, 07 Apr 2004 13:04:07 -0700
>X-Spam-Status: NO, hits=-4.90 required=5.00
>X-Spam-Level: Level 
>
>
>
>Serguei Krivov wrote:
>>
>>Bertram,
>>I certainly agree with you - the ontology graph does not tell the whole
>>story (as it does not intend to). So ontology versus "classified
>>ontology" (ontology where all possible subclass relations computed with
>>DL reasoner) will have different graphs. 
>>
>>However all these statements about graphs can be attributed to formulas
>>(textual representation) as well. Textual representation does not tell
>>the whole story, unless it is the output from a DL classifier.
>>
>>I do not want to say that we do not need textual representation- for me
>>personally it is more meaningful then a graph. I wanted only to say that
>>whatever can be said about  graphs can be said about formulas either.
>>Did I miss anything?
>
>This may be a bit off topic, but I believe that visualizing and 
>manipulating an ontology is very much user-dependent.
>
>If we consider a tool that lets a user just browse, and possibly 
>interactively select concepts/properties of interest, my opinion is to 
>show the "classified" version.  Note that a classifier typically 
>computes direct super and sub classes (i.e., it generates a lattice: a 
>partially ordered set where every class has a least-upper bound and a 
>greatest-lower bound), whereby it is possible to compute all 
>subsumers/subsumees, etc. Displaying the lattice is more like displaying 
>the UML object diagram, which I think is much more benefitial to such a 
>user. (The exception to this is union defs -- not sure how that would 
>look.)
>
>Shawn
>
>>
>>Ciao,
>>serguei
>>
>>
>>
>>_______________________________________________
>>seek-kr-sms mailing list
>>seek-kr-sms at ecoinformatics.org
>>http://www.ecoinformatics.org/mailman/listinfo/seek-kr-sms