Re: freebase parallax: user interface for browsing graphs of data

Hi Georgi,

Georgi Kobilarov wrote:
> Good point about the perception of browsing as only going forward.
> When I started my work on UIs for graph data at HP Labs, the paradigm I
> wanted to apply was zooming and panning. I've never really solved that,
> but I still think there is an analogy there to pivoting. 
>
> As you said correctly, I'm trying to build more of a query builder than
> a browser. Zooming is an increase in the level of detail, and I think
> pivoting (in my bi-directional query building style of interface) can be
> seen as exactly that: Increasing the level of detail in a particular
> dimension for a set of resources. However, one could also interpret
> pivoting as panning because each set of resources (after several pivot
> operations) has the same level of detail. I'm unable to make a clear
> distinction here.
>
> In order to not confuse the user when applying the query builder
> paradigm, it is key to represent the query path accordingly. Where am I
> coming from and how do my previous choices influence the current set of
> results?
>
> Please have a look at [1] to see what I was thinking of to achieve that
> (in particular slide 18 ff.). Please note that this presentation shows a
> more advanced design of Humboldt, while my paper described an earlier
> prototype.
>   
I love slide #4 :)

Here's an alternative to presenting the query path:
    
http://people.csail.mit.edu/dfhuynh/research/ideas/zoomable-faceted-browsing/page-02c.png
    
http://people.csail.mit.edu/dfhuynh/research/ideas/zoomable-faceted-browsing/page-03c.png
    
http://people.csail.mit.edu/dfhuynh/research/ideas/zoomable-faceted-browsing/page-04c.png
Properties that have deep constraints appear like "finger tabs". The 
whole UI is meant to slide/pan from right to left as you "zoom" into a 
property.

This won't be so helpful, but let me say it anyhow: Is there another way 
that you can re-frame the problem? The problem here is building a query 
where a constraint might be not on an immediate property but a property 
several hops away. You're using pivoting to get "there"--to get over 
several hops. I've tried to use "group by" to achieve the same effect to 
some extent. Maybe there are other ways.

>> This is actually a very subtle point
>>     
> It is, but it's a good one. However, in my opinion, it is the result of
> a certain UI design where there are no transitions between two sets of
> resources. Animations could help?!
>   
Not sure if I understand...

> I like "tributary links". Actually, what those links represent are
> inverse facet values. I very much liked your way of highlighting all
> connected resources per facet value in NFB. Maybe that's the kind of
> interaction I was thinking of for the tributary links.
>   
Oh, you mean the yellow synchronized highlighting...

> Maybe there's a chance to collaborate here, what do you think? 
>
> I'm also not aware of any upcoming semweb UI workshops. WWW2009 might be
> a good place...
>   
We can carry on this conversation off the list.

David

Received on Friday, 22 August 2008 06:50:58 UTC