Log in
Register
Give us Feedback
InviteGroup: CWA working group 2.6: Nanopublications
The Concept Web Alliance is an organization founded to promote the semantically rich publication and dissemination of life science data and knowledge.
The basic underlying units of this knowledge will be 'triples' - concept-relationship-concept - represented using the Resource Description Framework (RDF) specification of the W3C. Technologies that are built around RDF such as RDF Triplestores, URIs, Query Languages and OWL will have immediate importance to this and other CWA Working Groups including 2.3 (Storage and Maintenance), 2.4 (Unique and Persistent ID) and 2.9 (Triple Browser and Reasoning). Whilst RDF Triples will form the core of CWA recommendations, data and knowledge have structural complexities that cannot be represented with triples alone.
This CWA Working Group is responsible for recommending the higher order structures amongst triples and the vocabularies for describing these structures. The resulting triple model will allow features such as domain context for a triple (e.g. gene expression under certain conditions), metadata for triples (e.g. authorship).
Created at: Wednesday 12 August 2009 @ 14:12:31 (GMT)
Members
Owner
24 members
1 shared item
1 announcements
This group has been credited 0 times
Log in to add Tags
-
Nanopublication / Format Working Group Mailing List
more than 1 year ago by Andrew Gibson
[ See All ]
No news
No credits yet
Linked Data
Non-Information Resource URI: http://www.myexperiment.org/groups/192
Alternative Formats
Copyright © 2007 - 2011 The University of Manchester and University of Southampton
Log in to make a comment
Dear group members!
Here I've started the initiative to make a formal comparison of triple stores.
As I physically cannot make an overview of all triple stores in the world and you happen to possess any interesting experience with some triple store, please, share it in a wiki (and mark with an appropriate category tag).
Dear group members,
Is the RDF triple model the only triple model under consideration or is it possible that this group will decide on a different model? It seems to me that RDF has its limitations. As the developer of WikiData I have tried to find out whether its triple model is compatible with RDF and I have written up some of the limitations of RDF I discovered. Please look at www.triplestore.org/Article/91 to see my findings. If anyone disagrees with my findings they can comment on the article as they like. Also, I have tried to give an overview of the key benefits of WikiData at www.triplestore.org/Article/90 which might be interesting to members of this group as well.
Regards,
Peter-Jan
Dear Peter-Jan,
Thanks for your comments / questions, I will try and summarize a response here to the points listed on your website. In trying to be brief, my answers may not be thorough.
1) Triples do not have identifiers.
Technically this is true, in so far that there is no explicit reccomendation for this. You are right that the reification model that makes it into the RDF spec is widely regarded as not the best approach. For the answer to this you need to look outside the RDF spec for 'Named Graphs' which would essentially allow you to do what you propose. Named graphs did not make it into the original RDF spec, but if the W3C chooses to update RDF, I believe this is something that will recieve a lot of attention. Many triplestores offer a contextual / quad based approach which already allows you to get this functionality to some extent.
2) URIs are not opaque
I see this as a choice of the data publisher rather than something that and standard like RDF should enforce. URIs (or IRIs these days) are supposed to identify resources on the Web, and therefore it is recommended that you have HTTP URIs, and these will always have some sort of structure. If you wanted, you could go for the sort of purl.org approach, and have a registry of [baseURI + UUID]s that resolve to other more structured URIs. You can also have UUIDs internal to your KB and control how URIs are handled through the interface / API.
3) Textual predicates
Again this isnt really a problem with RDF, but more about how RDF is being used. If you have ever tried to write a SPARQL query, or try to look at some RDF, where everything has an opaque ID then youll find that you spend most of your time visually dereferencing the right predicates. I suggest that its a convenience that the URIs of predicates are textual and theres a good chance that as interfaces for interacting with RDF become more promenant, the stuff thats there for humans will become less important. Its not that long ago that people were encoding their classes and resources with textual IDs, but as more applications know how to look for rdfs:label, this practice has diminished (Though you may want to consider how I would talk about rdfs:label here if it was opaque :-)
4) SPARQL is not XML
Im curious about this point, as this requirement has never occured to me, but it sounded like the sort of thing someone would have done. I did some searching and found RDFQ which is a specification that I think would meet your requirements. Importantly, notice that it is possible to define the query language itself in RDF, which I think just shows off its flexibility really!
I hope this helps - overall I suspect that what you have in WikiData is some sort of syntactic transformation of RDF + some restrictions on how you use it - though I am not volunteering to prove this in any way!
Best regards,
Andrew
Hi All,
I wanted to introduce myself, my name is Paul Groth. I'm a researcher in the Knowledge Representation and Reasoning Group at the VU University Amsterdam. Since I heard about the concept web at e-Science back in December, I've been thinking about some ideas for the nanopublications format.
I've written those up and posted them under the shared items section (http://www.myexperiment.org/files/356). I wanted to see what you thought and maybe use those as a place to start some discussion.
Thanks!
Paul