invitation to explore the Getty Vocabs on ImageSnippets

[Margaret Warren]

Hello all!

I would like to extend an invitation for all to explore ImageSnippets. http://www.imagesnippets.com - ImageSnippets is in Beta, it is open and free to upload images and look up terms from both AAT and ULAN to apply to image descriptions. 

I'm very interested in feedback and would really enjoy meeting some Getty Vocab enthusiasts who would like a more personal introduction and/or would like to help with beta testing. We could possibly pay some beta testers under the right circumstances. 

With ImageSnippets, it is easy to enter an object value term and look it up against several vocabularies: dbPedia, AAT, ULAN, Yago and Geonames and Wikidata (as well as custom designed datasets) and see how the entities compare. In testing we have often found that the AAT provides an entity that intuitively works best with images - as compared with dbPedia for example..but in other cases, other datasets provide a better entity. It seems that no one dataset as of yet has all of the answers :-) . It is really interesting to compare the terms in actual use when you are looking at an image and deciding which entity matches more appropriately. 

Currenly we only have about 15k images in the system, but many of these are creative commons, public domain images from Flickr collections such as the British Library, Library of Congress, etc. which have been triple tagged with entities from all datasets including AAT. We are just now beginning to use the AAT terms more. 

We have an easy to use function for importing images - with their tags and other metadata from Flickr - into ImageSnippets, where the tags can then be turned into Triple Tags.

I would be happy to work with anyone who would like a personal demo or happy to answer questions at anytime. 

The system is in Beta and continually being developed, so, like any application in this state - there are occasional hiccups. 

Best, 

Margaret Warren

[Vladimir Alexiev]

Hi Margaret! I added this to our list of Usages on 12 June from your tweet, and amended the info from the above message.

[Margaret Warren]

Hi Vladimir, 

Thanks so much! Can you direct me to where that list is located? I have not been able to find it. 

Best,

Margaret

[Richard Light]

Margaret,

This is an interesting project. Creating and preserving image metadata is an issue which crops up regularly in the cultural heritage world, e.g. the recent discussion on the MCG list about embedding IPTC/EXIF data [1].

A few questions spring to mind:

Will it be possible for providers of images to extract all their Linked Data metadata, or will it remain 'locked in' to the IS system?

Are you planning to support Linked Data content negotiation, so that your image URLs could also deliver RDF/XML, Turtle, etc. in response to a suitable Accept header in the HTTP request?

Do you have plans to support external authorities by defining a web service interface, as a complement to the current option of hand-entering the whole framework?  This is presumably how you already interface to Geonames, DBpedia and AAT.

Thanks,

Richard

[1] https://www.jiscmail.ac.uk/cgi-bin/webadmin?A1=ind1506&L=mcg#33

--
You received this message because you are subscribed to the Google Groups "Getty Vocabularies as Linked Open Data" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gettyvocablo...@googlegroups.com.
To post to this group, send email to gettyv...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gettyvocablod/a373cc5b-7fa7-4e21-92a5-0062fea23c34%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--
Richard Light

[Margaret Warren]

Hi Richard,

Thanks so much for writing with questions. As an aside - we support all of the IPTC/EXIF fields including the newest standard. When you are in the application - just type IPTC in the plain text box and their sample image will come up and go to the details tab - all of these fields are IPTC or Dublin Core schema - but then click on the 'More' button at the bottom of this list and you will see the entire new schema and can select items to use. I guess I need to post something on this MCG list which I've never heard of before - what is that? 

The answer all of your other questions is basically 'yes' - you wrote: 

 

>>>>>>Will it be possible for providers of images to extract all their Linked Data metadata, or will it remain 'locked in' to the IS system?

 

answer:  yes, it already is open -- the idea of IS is to allow content creators or curators of collections - to assert their desired copyrights - even if Public Domain and the image metadata and rights openly published. All metadata in ImageSnippets is published daily to a triple store that is open, and queryable from our SPARQL endpoint as well as published on datahub.io. 

 

>>>>>> Are you planning to support Linked Data content negotiation, so that your image URLs could also deliver RDF/XML, Turtle, etc. in response to a suitable Accept header in the HTTP request?

 

answer:  yes we do support content negotiation  (aka 5-star linked data)  - see the first answer above for clarification. Additionally we are also continuing to expand on the options available for export as we speak - there is an export button that shows some of the options that currently work (simple web gallery, JSON, etc) but we are currently improving upon these. 

 

>>>>>> Do you have plans to support external authorities by defining a web service interface, as a complement to the current option of hand-entering the whole framework?  This is presumably how you already interface to Geonames, DBpedia and AAT.

 

answer:  if you click on the 'user datasets' button in the interface, you can see a list of options of datasets that a user can subscribe to, as well as building their own custom datasets - we currently either pull directly from the services (as in Geonames) or use a local copy of dbPedia (which is just faster and less problematic), but adding datasets is extremely easy for us - we added AAT and ULAN each within 24 hours of when they were each released (both of them are local copies on our server)  bottom line: we can build custom interfaces very quickly that add whatever dataset you would like to use.

NOTE --- (I am not sure if people need a little more guidance to see this at work), but to look up entities:  you type in a word for the 'object value' when you are building a triple - a small exclamation point will come up - if it is black, it has found matches in one, if not all subscribed datasets - CLICK on the black exclamation point and double click to choose the most appropriate match, or build multiple triples - one for each if desired. If the exclamation point comes up red - you have the option of creating an entity - but I caution that people should try another word or phrase before doing this, or before they become more familiar with the system. We are also currently working on dramatically improving this interface to make it much simpler to resolve the entities. 

Of course, while you are beta testing, you can add nonsensical entities if you want.

Very happy for your questions and I would be very happy to give a personalized demo. 

Best,

Margaret 

[Margaret Warren]

Hello again Richard and all,

I answered the following questions below about providing 5-star linked data with content negotiation via SPARQL and published on datahub.io already, but I also wanted to add the following use case as well. 

ImageSnippets also allow you to use your own hosting of your images and simply provide your image URLs, or even import your Flickr URL's into ImageSnippets if you have shared them, for example, in the Commons. So we don't need to 'hold' your data in the system. ImageSnippets is meant to function as a tool or a service (however you want to look at it)  for which you can create the metadata which is query-able and published daily. 

We are absolutely committed to not building another silo!  ImageSnippets is about provenance, open-ness and encouraging people (and organizations) to be able to publish their data with their copyright intentions clearly asserted. 

Thanks,

Margaret

On Thursday, July 2, 2015 at 3:18:25 AM UTC-5, Richard Light wrote: