Re: reworking welkin

From: Vineet Sinha <vineet_at_csail.mit.edu>
Date: Sun, 05 Feb 2006 05:27:06 -0500

Arvind Venkataramani wrote:
> Vineet Sinha wrote:
>> And another project, Relo: http://relo.csail.mit.edu/
>> ...
>>
>> Relo does allow answering the question of the type in your example,
>> i.e. ‘show me all authors who have published with penguin’. I'm
>> seriously considering adding a faceted refinement interface (similar
>> to longwell / magnet) and I'd be interested in helping build features
>> to answer other more complex questions, using visualization or more
>> traditional techniques.
>
> Relo provides visualization and manipulation capabilities for what
> are nicely structured domains - a set of classes inheriting from and
> coupled with each other make for a very neat graph. there's also the
> fact that the relationships between the nodes in the graph are of a
> limited finite type, which can then be coded for. I think Relo's a
> great example of domain specific RDF visualization, but what I want
> to do is something generic that isn't tied down to a specific
> representation.

My approach to building usable tools is to first start with doing a
good domain specific solution - which I believe that Relo is
approaching. The next question is whether any of these ideas can be
applied to other domains, and this is exactly what I plan on showing
with Relo in the next month or two.

Relo can be thought of as Welkin with support for stylesheet's added on
it. This stylesheet can control 4 families of parameters:
o Appearance: Similar to what is being considered for Welkin, i.e.
different node types having different shapes and colors.
o Visual constraints: The ability to map the RDF relationsships to
represent horizontal left-to-right constraints, vertical constraints, or
event containment information.
o Agents: To deterministically but automatically add nodes depending on
the criteria.
o Exploration Relationships: Which relationships should be made
available easily for exploration, either via context menus, or
navigation aids (hovering buttons).

Currently support for multiple stylesheet's is already been made
available in the code, and I essentially need to provide a non-java
provided stylesheet input engine.

So, without providing such a stylesheet, Relo should provide exploration
capabilities similar to Welkin with support for exploration based
showing of partial graphs. When additional information is provided
usability should be better.


>> ...Relo is designed to look at *parts* of *large* graphs
>> of information. Visualization techniques have shown to work poorly
>> showing large graphs, so we are spending time figuring out how to
>> select portions of the graphs to show to the user.
> This is where I differ - I'm not building a graph visualization tool,
> which emphasises node objects and spatial relationships. Not everything
> that is expressible as an RDF graph *is* actually a graph, nor in many
> cases is a graph representation the best way to understand RDF, so I'd
> rather be able to look at RDF as both multivariate data and as graph data.

Yes, the Relo visualization engine focuses on showing the graph parts of
the data. Relo is built using the Eclipse plug-in mechanism, and at the
core is basically RDF, so if you want to go the Relo way, it should be
fairly easy to build an extensions focusing on the multivariate parts.
This is precisely how I plan on adding support for the faceted
refinement interface.


>> If you have chance, I'd be interested in hearing a citique of Relo.
> If I can get it to run, that is :) When I open a Relo session with a
> project or two opened, i get a dialog box asking me to select a path to
> open, with no paths to open. I'm not sure what I should be doing with
> it, but we can take that discussion offline.

It seems that the site might not have been clear on the startup
directions, to make thing clearer I have just added a set of screencasts
on the site: http://relo.csail.mit.edu/screens.html

If you are still having problems, lets discuss it on:
relo-users_at_csail.mit.edu


Vineet
Received on Sun Feb 05 2006 - 10:26:05 EST

This archive was generated by hypermail 2.3.0 : Thu Aug 09 2012 - 16:39:18 EDT