intuitive interfaces

Hi All,

I have been having some good successes with the ImageSnippets 
(http://www.imagesnippets.com) interface - designed for non-RDF techs.

Since my software is about describing images with linked data, I get a 
mix of users and institutions who generally know nothing about linked 
data or RDF.

A number of users can use the software without any training. But I can 
train users to use the basics of our relations in about 5 minutes and 
many of them become quite proficient at becoming 'triple-taggers' after 
using the system for about 15 minutes. We keep making it quicker and 
easier for users to look up entities in existing datasets like:  
DBpedia, Yago and Getty's Art & Architecture Thesaurus. Additionally, we 
use DBpedia's auto-entity extractor to find terms easily that users can 
map with a relation as part of a triple and we can even consult Clarifai 
and Cloudsight to see how they 'see' the image and then allow users to 
map those terms with relations in triples as well.

In one of my on-going research projects with IS, I have upload groups of 
images based on collections of 3 ambiguous keyword terms from the 
yahoo-flickr collection (i.e. :  train/railroad/bridge --- or: 
water/boat/island)  and I have a team of  'triple taggers', who review 
these images and build triple tags from looking at the photos and 
deciding what relations will work best with the keywords.

In another research project, we are working with subject matter experts 
to use the tool for ontology engineering from looking at photos.

This requires an advanced level of triple-tagging that is taught to the 
triple-taggers as a 'level 2' since we can also create new types and 
entities within the tool when items don't exist in an existing dataset - 
and it requires a little more skill to know how and when this should be 
done.

In a 3rd project, we now have a word-press plug-in gallery that allows 
you to design dynamic web galleries pulling SPARQL queries on a schedule 
or on-demand from the RDF data in ImageSnippets. 
http://www.margaretwarren.us/junkyard-art is an example.  - All of the 
metadata is pulled from the triple-store and included on the remote site 
as JSON-LD as well as linked back to the original html + RDFa file.

All of this is done relatively affordably and for the most part, 
consistently and produces some fairly clean data.

Margaret Warren



On 4/10/2018 8:51 AM, Bram Biesbrouck wrote:
> Hi all,
>
> This is exactly my thought as well: Linked Data needs a bit more 
> intuitive and user friendly software.
> That's why we started the open-source Stralo project (www.stralo.com 
> <http://www.stralo.com>) a few years back.
> The goal is to hide away the complexities of working with triples as 
> much as possible and let people create fun stuff with Linked Data 
> instead of having to learn it.
>
> Take a look, we're always looking for extra interested developers to 
> share the love with.
>
> all the best,
>
> b.
>
> <http://www.reinvention.be>  *We do video technology*
> Visit our new website! <http://www.reinvention.be>  *Bram Biesbrouck*
> bram.biesbrouck@reinvention.be <mailto:bram.biesbrouck@reinvention.be>
> +32 486 118280 <tel:0032%20486%20118280>
>
>
> On Tue, Apr 10, 2018 at 2:09 PM, Mikel Egaña Aranguren 
> <mikel.egana.aranguren@gmail.com 
> <mailto:mikel.egana.aranguren@gmail.com>> wrote:
>
>
>
>     2018-04-10 12:51 GMT+02:00 Ruben Verborgh <Ruben.Verborgh@ugent.be
>     <mailto:Ruben.Verborgh@ugent.be>>:
>
>         Hi Krzysztof,
>
>
>     Hi, congrats!
>
>
>         Congrats!
>
>         > I am very glad to report that IOS Press went all in on
>         Linked Data as backbone for their internal data management and
>         sharing.
>
>         Any details on how you are doing this?
>         Do you have a CMS on top of a triplestore?
>
>
>     I'm really interested on this too. I think the only way for Linked
>     Data to succeed is to be consumed by the internal programs of the
>     entity (Company, Government) that publishes the data. Otherwise,
>     the effort needed to build proper Linked Data (Data Quality, URIs,
>     ...) is too much to be worthy if publication only is involved.
>
>     Cheers
>
>
>         Best,
>
>         Ruben
>
>
>
>
>     -- 
>     Mikel Egaña Aranguren, Ph.D.
>
>     https://mikel-egana-aranguren.github.io
>     <https://mikel-egana-aranguren.github.io>
>
>
>