[open-linguistics] new colored LLOD diagram

Sat Sep 21 22:30:26 UTC 2013

Hi Tom,

> What's the criteria for inclusion?  It seems weird that DBpedia  
> qualifies,
> but Freebase doesn't when it's got everything DBpedia does plus Wordnet,
> data from the Long Now Foundation's Rosetta Project and other linguistic
> resources.

We had a brief discussion about criteria, without a clear result.* And  
right now, the criterion is just a pragmatic check of datahub entries and  
whether these are tagged for the group (except for obvious spam, and  
non-RDF data sets, look in the script). You're right, freebase could be  
included, as well. It only needs to be tagged as such on datahub.

* Personally, I think that DBpedia is somewhat borderline, too, as it  
contains very little specifically linguistic information, but it qualifies  
through its application in NLP and through its multi-lingual character:  
Even if originally not intended to be, the linking between different  
language versions yields something like a multilingual wordlist (which  
would be a linguistic resource in any case).

> Other things that caught my eye:
> - there seem to a lot of free-floating circles for a *linked* data  
> diagram.
>  Is this an issue with missing data or are these data sets really islands
> that are only linked internally?

Actually, the diagram is drawn from the datahub entries, and we did not  
verify the information given there. This is also true for the links. The  
metadata there is incomplete, and several times, I know that links should  
exist, but are not documented in the metadata, e.g., for the "simple"  
resources (cf. the MLOD submission:  
http://www.semantic-web-journal.net/content/parolesimple-%E2%80%98lexinfo%E2%80%99-ontology-and-lexicons),  
or for OLiA and lemonuby. Some are real islands, but if so, they should be  
linkable. John and me discussed whether these should be excluded, but as  
some of them are actually a concrete result of collaborations within the  
group whose linking was just not documented yet (think of simple), I  
thought we should keep them in there. The diagram built at MLODE-2012  
followed the same policy.
Of course, a better solution would be to be stricter and to work our way  
through the metadata. And this is something that really needs to be done,  
actually ...

> - DBpedia Live is redundant with DBpedia (English)

Possibly, but apparently, there are two datahub entries. Anything from the  
DBpedia folks on this ?

> - if the diagram is going to exist in an online version, it'd be cool to
> link the circles back to their CKAN entries (or their home pages)

Sure. In SVG, this is possible and we had this in the last hand-written  
diagram (Feb 2012). Right now, we generate GraphML and I haven't checked  
whether it supports hyperlinks.

Best,
Christian