[Dev] SMC Browser - update

[Matej Durco]

Dear Oddrun,

thank you for the positive feedback.
I also have many ideas in my head, where this could evolve.
One I am currently intensively working on, is to weave in instance 
information,
i.e. to be able to see, which profiles and components are used an how 
often..

Regarding the problem of visibility of the full graph:
True, this is not solved optimally yet.
First, I usually try to play around with the parameters:
lower link-distance and lower charge squeeze the graph together.
When resizing the window, the scroll-bars adjust - sometimes,
unfortunately this is currently inconsistent and rather frustrating.
Third possibility is to Zoom out the browser (Ctrl -).
Funnily, once you get the whole graph into the picture, when zooming out,
the scroll-bars seem to adjust correctly, when you then go back to 
normal view (Ctrl 0).
Then there is always the possibility of exporting the graph.
By clicking on the download link, you get the pure svg opened in a 
separate tab,
where provisions where made, to display the full graph.
You can store this as svg, however you cant interact anymore with the graph.
Depending on the situation, you could also try to traverse the graph,
meaning you select the last nodes visible, and deselect the originally 
selected ones.
This way, the graph should move to the left.
And final possibility: bigger screen ;)

Hope this helps a bit.

Matej



Am 28.08.2013 10:08, schrieb Oddrun Ohren:
>
> Hi Matej,
>
> Thanks for the new version!  As I have said before, I find your tool 
> highly informative and helpful. It also visualizes very clearly some 
> of the problems with CR as it is today, notably the many duplicate or 
> near-duplicate components and elements. I also think that a controlled 
> vocabulary for domains (of profiles) could make the profile-domains 
> graph really helpful. Introducing a similar attribute for components 
> (something like "described object" with an accompanying 
> vocabulary/ontology) would make the foundation for a corresponding 
> graph /component-described object./ That way, it would be easy to 
> detect components which describe persons, for example, without having 
> to rely on them being called "person-" something or other.
>
> With further development, I think CMD Browser could become a _CMD 
> analysis and management tool_. For this we would need facilities like
>
> ·profile/component comparison (e.g. compare selected components in 
> terms of overlap, inclusion, etc.  Relations to investigate must of 
> course be clearly defined), with textual/tabular report
>
> ·detect profiles/components/elements which are similar to a given 
> component ("similar" defined partly by user, e.g. "includes >= 80 % of 
> the same DCs")
>
> ·...and more. We would have to think more about this...
>
> Even further down the line, perhaps a _graphical modeling tool _could 
> evolve, allowing the metadata modeler to create/update profiles by 
> dragging/dropping components, elements and data categories on to a 
> graphical canvas.
>
> _Back to the current version:_Using Chrome, I have a little trouble 
> with graphs extending past the window borders. The scroll bars don't 
> seem to cover the whole canvas. Any attempt to manually move the nodes 
> is undone by the automatic layout mechanism. I tried freeze, but then 
> I still can't get hold of the nodes outside the accessible part of the 
> canvas. Anything else I could have done, apart from using the layout 
> parameters in the top bar?).
>
> Best regards,
> Oddrun
>
> -----Opprinnelig melding-----
> Fra: Matej Durco [mailto:xnrn at gmx.net]
> Sendt: 27. august 2013 19:42
> Til: CLARIN EU developers; Oddrun Ohren
> Emne: SMC Browser - update
>
> Dear all,
>
> there is an updated version of the SMC Browser available (under the old
>
> address):
>
> http://clarin.aac.ac.at/smc-browser/
>
> Besides freshly fetched data, there are also some new features and fixes:
>
> * profiles that are still private, but already have instance data 
> exposed have been also processed, making it altogether 154 available 
> profiles.
>
> * introduced new type of nodes: *profile-groups*
>
>    based on profile metadata: creatorName, groupName and domainName
>
>    virtual nodes (for each distinct value of the three fields) are 
> being generated and linked with appropriate profiles.
>
> so now it is easy to see which profiles are grouped together or who 
> created which profile (i hope the authors dont mind being dots in a 
> graph - aren't we all just tiny dots in a giant graph after all ;) )
>
>   especially there is a dedicated group for profiles, that are not
>
>   public yet: "unknown-profile"
>
>   you can inspect directly under:
>
> http://clarin.aac.ac.at/smc-browser?graph=/smc/data/smc-graph-groups-profiles-datcats-rr.js&depth-before=1&depth-after=1&link-distance=120&charge=594&friction=75&node-size=4&labels=show&curve=straight&layout=horizontal-tree&selected=profile_group_unknown_profile&
>
> * it is now possible to select different graphs (param: graph)
>
>    currently 4 different ones are available:
>
>    1. the basic one (profiles -> components -> elements -> datcats)
>
>    2. additionally rendering the new profile-groups and relations 
> between data categories (from Relation registry)
>
>    3. just rendering profiles and data categories (with direct links 
> between those, skipping all components and elements
>
>    4. as 3. but again with profiles-groups and relations
>
> 3. and 4. are good to better inspect profile-proximity
>
> * all data categories should be resolved now.
>
>    (in the previous version many datcats had no definitions fetched from
>
> isocat)
>
>    I also encountered quite a few concept-links with leading space, 
> and other inconsistencies. I will soon provide a list of the affected 
> CMD-elements, perhaps, we can clean this up in the Component registry...
>
> * the (statically generated) dataset (especially the statistics) are 
> getting quite big, so it may take a bit until everything is loaded.
>
> a more dynamic version sending around smaller chunks of data is 
> scheduled. meanwhile I have to ask you for patience.
>
> I hope you like it.
>
> looking forward to feedback.
>
> best,
>
> matej
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clarin.eu/pipermail/dev/attachments/20130828/78397e2e/attachment.html>