Re: [Myexperiment-discuss] Scenario for editing and linking within myExperiment Packs

[David R Newman]

Hi Robert,

Don and I have considering how best to do this, based on general ideas
that have been floating around in the E-Lab TAG.  I have quickly
digitized our scribblings at
http://rdf.myexperiment.org/img/ROs_ORE_SKOS.png

Basically a pack uses ore's aggregate property to aggregate the actual
files.  We then use pack entries that are ORE proxies to refer to them
in the context of the pack.  At this point we can build "relationships"
between these pack entries using a SKOS taxonomy to describe these
relationships.  This would allow different communities to define their
own taxonomy of relationships that may well build on a central offering,
if there is agreement on certain core relationships.

Looking further into the future I see UI as being one of the most tricky
problems. Your use case in some ways negates this by having very
specific relationships that a user must fill in as they upload / add
things to a pack.  However, I see there being a need to retroactively
create links.  How do you present this to a user so they can create
these links. Some graphical "click the two things you want to link, then
specify the type of relationship" interface would be way to go but it is
critical to ensure it is reliable cross-browser and supports the user's
preference of relationship concepts.

Regards

David Newman

On Thu, 2010-09-09 at 14:04 +0100, Robert Stevens wrote:
> all
> 
> After a discussion with DDR on a recent visit to Manchester I've put
> together a scenario for the editing, tagging and linking of elements
> within a myExperiment pack. The scenario is  below and, as usual,
> comments and quesitons welcome.
> 
> robert.
> 
> e-LICO (http://www.e-lico.eu/) is an EU project for a data-mining
> eLab. Taverna is used to make data-mining workflows that use
> services/operators from a tool called rapidMiner
> (http://rapid-i.com/content/view/181/190/) . Part of e-LICO is the
> planning of workflows and `meta-learning' over metadata collected
> about runs of the workflow and the service/operators used in the
> workflow in order to optimise the planning process. e-LICO wishes to
> use myExperiment both as a forum for collecting data-mining workflows;
> input to those workflows; and various data about the workflow run --
> the `experiment'.
> 
> 
> An e-LICO user runs a data-mining experiment and wishes to upload its
> workflow and links to its data as a pack to myexperiment. She has run
> the experiment within Taverna, which itself talks to RapidMiner, and
> has viewed her results. the various data (input, metadata, and so on)
> are stored on RapidMiner and the URL are available. 
> 
> She uploads her workflow to myExperiment via the myExperiment plugin.
> she next presses the `upload experiment button. she is offered:
> 1. A link to her input data;
> 2.   A link to the metadata about the workflow and its data
> 3. A link to the description of goals and task;
> 4. A link to some provenance for the workflow run.
> she can say yes/no to these links. 
> 
> 
> tags, such as `input', `metadata', `goals', `provenance, are added to
> these links. In addition, a link from the input data to the port in
> the workflow is made.
> 
> Our user then uploads all this stuff to myExperiment. The URL are then
> visited and the data stored somewhere or other.
> 
> If the workflow already exists on myExperiment, then the e-LICO
> workflow run pack can be linked directly to  this pre-existing
> workflow. Similarly, the input data, if used elsewhere as input for
> either this workflow or another workflow, could be linked.
> 
> the e-LICO meta-learners periodically harvest e-LICO experiments from
> myExperiment. they gather the various input and meta- data for
> analysis for eventual optimisation of workflow planning.
> 
> 
> 
> 
> 
> 
> Robert Stevens
> Reader in BioHealth Informatics
> School of Computer Science
> The University of Manchester
> Oxford Road
> Manchester
> M13 9PL
> 
> address@hidden
> http://www.cs.man.ac.uk/~stevensr
> 
> 
> _______________________________________________
> Myexperiment-discuss mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/myexperiment-discuss