RDF/DataId about gerbil

[RicardoUsbeck]

we will use json-ld embedded in the html. Please remember to use a branch and git pull mannor to hand-over the source code.

[MichaelRoeder]

The first version looks good. :)

But are you sure that this is correct?
"annotator" : "http://gerbil.aksw.org/ns#http://gerbil.aksw.org/ns#annotator/dataid.ttl"

[der-bruemmer]

URIs depend on the namespaces we want to choose. I would just drop the files into the webResources folder. So URIs should be

http://gerbil.aksw.org/gerbil/webResources/corpora/N3-RSS-500 for the corpora and
http://gerbil.aksw.org/gerbil/webResources/annotators/AGDISTIS for the annotators.

Vocabulary namespace should then be

http://gerbil.aksw.org/gerbil/webResources/vocab#

That way, you don't have to change server configuration. Would that be ok @MichaelRoeder ?

@cirola2000 can you make the changes?

[MichaelRoeder]

This is a good idea. However, I would prefer to use something like
http://gerbil.aksw.org/gerbil/dataId/corpora/N3-RSS-500.ttl for the corpora and
http://gerbil.aksw.org/gerbil/dataId/annotators/AGDISTIS.ttl for the annotators,
because webResource sounds a little bit inappropriate. I think, I can add the dataId to the routing.

The same holds for the Vocabulary. I think we could use http://gerbil.aksw.org/gerbil/vocab#

[MichaelRoeder]

I Added a mapping for the dataId files. Now, you can add the files und src/main/resource/dataId/corpora or src/main/resource/dataId/annotators. I added two files that are accessible by the URIs http://gerbil.aksw.org/gerbil/dataId/corpora/test.txt and http://gerbil.aksw.org/gerbil/dataId/annotators/test.txt (They are commited to the master branch but not on the running server).

But I'm not sure what you need to refere to parts of the vocabulary. Are these files, too?

[MichaelRoeder]

Hi,

I took a closer look into the DataIDGenerator, because it creates an exception if I try to open an experiment which does not exist. (Well, normally nobody should do this, but I wanted to know where the excpetion came from)

For better understanding it I structured the class and made some minor changes.

• I didn't really understand why everything has been created as Properties.
• I think we should discuss how an experiment is defined and with which Resource the results (F1, precision, recall) should be connected.
• I didn't really understand gerbil:dsd. What does it mean?
• I think it is not possible to give the experiment resource a URI like http://gerbil.aksw.org/gerbil/experiment&id=201411070000 ?

Maybe I just got an unfinished version of the DataIDGenerator and I hope it wasn't cheeky that I made changes to your class without discussing these issues. Because I didn't wanted to override your class, I pushed my changes into a separate branch. Maybe you want to look at it: https://github.com/AKSW/gerbil/tree/dataId-temp

[cirola2000]

@MichaelRoeder
The first version of the dataid class was very straightforward since it was made before @der-bruemmer create the ontology. Hence on Thursday I just changed the properties names. I was planning for today creating vocabulary classes and restructuring the generator class.
I notice that you have done a good job in the dataidtmp branch and thanks for that. Feel free to marge it with the master branch!

About the experiment URI we definitely can change it for http://gerbil.aksw.org/gerbil/experiment&id=201411070000 .

About the gerbil:dsd maybe @der-bruemmer could answer us.

[der-bruemmer]

@MichaelRoeder gerbil:dsd is a concept defined by the RDF Data cube vocabulary. Basically, it specifies the structure of the dataset (the dimensions, measures and so on that can be found in every observation).

The vocabulary is only a single file, but maybe we could have http://gerbil.aksw.org/gerbil/vocab/gerbil# or something like that to keep the routing strategy?

[MichaelRoeder]

@cirola2000 merge is done.

[MichaelRoeder]

@der-bruemmer I couldn't find a good solution for creating a mapping from http://gerbil.aksw.org/gerbil/vocab#* to a single file. I added some methods to the MainController class that simply load the files from the classpath and return them as String depending on the accept header of the request.
application/json+ld --> vocab/gerbil.json
application/json --> vocab/gerbil.json
application/rdf+xml --> vocab/gerbil.rdf
text/turtle --> vocab/gerbil.ttl
text/plain --> vocab/gerbil.ttl

Is this the mapping you wanted to have?

[cirola2000]

@MichaelRoeder I changed the URI to:
http://gerbil.aksw.org/gerbil/experiment?id=201411090001#experiment_201411090001_task_0
The base URL (http://gerbil.aksw.org/gerbil) is automatically created.

@der-bruemmer the dataid URIs now looks like:
http://gerbil.aksw.org/gerbil/dataId/corpora/DBpediaSpotlight.ttl
http://gerbil.aksw.org/gerbil/dataId/corpora/AIDA_CoNLL.ttl

What do you think?

[der-bruemmer]

@cirola2000 can you map all AIDA_CoNLL (-Training; -Test etc) corpora to just AIDA_CoNLL ? Can you do the same for Microposts? We don't really have a way to describe the differences between these corpora yet, so they should link the same description (imho).

Depending on @MichaelRoeder's reply regarding content negotiation, can you remove the file format from the URIs? Like this: http://gerbil.aksw.org/gerbil/dataId/corpora/DBpedia_Spotlight

@MichaelRoeder Great! Yes, content negotiation is very nice. Can you add a rule to forward html requests to rdf-xml? Is there content negotiation for the corpora and annotator DataID files, too?