DBTune blog - Comments

HENRY: A small N3 parser/reasoner for SWI-Prolog - ana

ana — Wed, 14 May 2008 02:14:26 +0200

Great tips. Thanks.

Linked Data on the Web 2008 - Yves

Yves — Mon, 12 May 2008 18:12:08 +0200

Hello Paul!

I actually completely agree with you. Explicit licensing in a machine-readable format is the key (and Creative Commons is the perfect example proving it). And the more open the licensing is, the better :-)
It is also a daily problem that the "re-user" of data have to face as well. It is clearly not usual for data providers to give explicit licensing (apart from some exceptions, like Musicbrainz) - in the best case they provide buggy licensing that don't really apply to data.
We must ask not only for the data to the providers, but also for explicit, machine-readable, licensing of it. And as you say, it needs a real meeting of minds between the three communities.
But I have the feeling (at least in the domain I am mainly interested in) that most data providers don't even realise there is a problem here: only license drafters and re-users do. So we *really* need a serious outreach there.

It might be useful to look at how such a realisation occurred in the Creative Commons and in the open-source worlds. In the CC world, I guess it is mainly due to the numerous court cases, and a sudden realisation that traditional licensing schemes were not fitted to handle new creative processes. In the open source world, I really have no idea :-)

I just hope we won't need court cases to make such a realisation occur...

Linked Data on the Web 2008 - Paul Miller

Paul Miller — Mon, 12 May 2008 12:48:14 +0200

Yves

yes, it was a good workshop, wasn't it? :-) I hope you had a good time touring China. Other than the wandering (almost) Dr Heath, we just had to make do with two crammed days immediately after the conference, but managed to fit quite a bit in.

To respond, specifically, to your point about Linked Data... yes, we definitely do need to be (and we are) talking to the data owners themselves.

We also need to be talking to those who re-integrate, recombine and reuse the data (such as yourself). Firstly, in the eyes of the law 'ignorance is no defence.' Just because you can re-use someone's data doesn't mean you're allowed to. Those who reuse data need to be aware of the issues around what they're doing... especially as early 'experiments' and 'demonstrators' grow to become services that users rely upon; and that the data owners might actually notice and (possibly) question.

Secondly, we all need to be applying ourselves to understanding ways in which a whole raft of licensing terms can be better expressed in machine-actionable form. I'd prefer that users didn't need to engage with most of this stuff. It would be far better if we had data stores that expressed their licensing terms, and applications that obeyed those terms as they worked with data coming from diverse sources. That requires a real meeting of minds between license drafters, data owners, and application builders. Several starts have been made, but we're certainly not there yet.

We all, I think, want data to be as freely and widely available for re-use as possible. Licenses such as Open Data Commons aren't intended to constrain use. They're intended to explicitly and proactively describe all the things that you can do with someone else's data. They're an important missing piece in the journey from innovative and exciting experimentation (where, if we're honest, 'rights' and 'ownership' are rarely as carefully acknowledged or respected as they should be) to robust and sustainable delivery of services and applications.

MySpace RDF service - madrid

madrid — Sat, 03 May 2008 17:33:28 +0200

Great post. Saludos desde madrid (Hello from Madrid)

Yay for SemanticCamp! - Yves

Yves — Tue, 15 Apr 2008 00:24:32 +0200

Hi Chris!
Nice! I just registered to the mailing list as well - this new specification (and also the future time-related work) looks just great... :-)
y

Yay for SemanticCamp! - adrideo

adrideo — Mon, 14 Apr 2008 23:27:27 +0200

Thought I'd drop a link to the updated URIplay ontology, which was an output of the Sunday morning audio and video session:

http://uriplay.org/spec/ontology/

Many thanks to all who helped us.

Playing with Linked Data, Jamendo, Geonames, Slashfacet and Songbird - Roberto

Roberto — Sun, 13 Apr 2008 21:58:22 +0200

Great! It's the first attempt to build a so rich mashup. I'll investigate geonames too to implement on http://www.qmpeople.com/

13.1 billion triples - Yves

Yves — Sun, 06 Apr 2008 00:54:11 +0200

Thanks a lot for the kind words and the great post, Amy!!
:-)

13.1 billion triples - Amy Stephen

Amy Stephen — Sat, 05 Apr 2008 06:18:17 +0200

Very cool - I responded "The Potential of Great Music, CC Licensing and 13.1 Billion Triples" http://opensourcecommunity.org/2008... Thanks for your work!

13.1 billion triples - Beqn

Beqn — Thu, 03 Apr 2008 19:04:10 +0200

well, I'm not entirely sure, but I had thought it would allow access to structured public profiles, but only when you have a developer authentication key. So there is some access restriction. I'm not entirely sure what the limits are from this perspective, though I imagine we'll get it sorted out quite easily on Sunday.

13.1 billion triples - Yves

Yves — Thu, 03 Apr 2008 12:07:29 +0200

But if I get OpenSocial right, it provides you with data when you're logged in (eg. a MySpace widget wanting to access your list of friends)? Does it provides a way to access a more structured version of the public profiles?

13.1 billion triples - Beqn

Beqn — Thu, 03 Apr 2008 02:19:42 +0200

hmmm... Well I can't help with the sparql end points for myspace, but hopefully shortly after this weekend's hackfest (http://opensocialapis.blogspot.com/...), we'll have xml feeds out of myspace...

The Quest for Canonical Spelling in music metadata - Yves

Yves — Sat, 29 Mar 2008 14:16:47 +0100

Hi Daniel!
It can, to handle a different label per language - but I don't think it can help for the merging of the different URIs, and also for cases where you need to handle several labels in one single language.
I guess my point was more pro-merging of all the different URIs for a single artist/track/album (the ID you get with the fingerprinting lookup software corresponds to the fingerprint). Then, all the different alternate spellings can be attached to it, and picking one up that you consider as canonical is another problem (slightly less important, in my opinion).

On a side-note, it'd also be neat if the voting system handled voting for a label in a particular language :-)
Cheers!
y

The Quest for Canonical Spelling in music metadata - Daniel O'Connor

Daniel O'Connor — Sat, 29 Mar 2008 01:55:11 +0100

Doesn't xml:lang solve some of the problem here?

The Quest for Canonical Spelling in music metadata - Norman

Norman — Thu, 27 Mar 2008 21:35:44 +0100

alternativelY

The Quest for Canonical Spelling in music metadata - Norman

Norman — Thu, 27 Mar 2008 21:35:00 +0100

Hi Yves!
Or, alternative use the option -url instead of -nometadata.. ;-)

The Quest for Canonical Spelling in music metadata - Yves

Yves — Thu, 27 Mar 2008 18:09:42 +0100

Actually, I just figured out that:
http://ws.audioscrobbler.com/finger...
gives access to the results corresponding to <id>.

It then looks like <id> identifies the actual fingerprint within your db. So I guess the merging you're planning will "propagate" this identification to further resources (artists, albums, etc.).

Cheers, and thanks again for the hint!
y

The Quest for Canonical Spelling in music metadata - Yves

Yves — Thu, 27 Mar 2008 17:41:39 +0100

Hi Norman!

Oh, cool, it actually finds an ID with -nometadata: 281948 for the same example as above. Is there a way to associate this ID with an URI that I can dereference to get all information about this ID within your database (including all alternate labels)? A quick look at the AudioScrobbler API seems to key such a request on the artist name.

And that's really good news for the merging!! Also, do you have any plans on providing some structured data from the artist/track/album URI (RDF, RDFa or microformats) ?

Cheers, and thanks :-)
y

The Quest for Canonical Spelling in music metadata - Norman

Norman — Thu, 27 Mar 2008 16:07:57 +0100

It is true that currently we have 21 different artist pages for each (mi)spelling but our idea (thanks also to the fingerprint technology) is to merge them into a single one. The identifier you're talking about is clearly central. Indeed you can get such ID it if you specify -nometadata as option to our program! :)

As for different languages, our goal is to return whatever is specified in the preferences of the user. If he/she is japanese we will return the kanji spelling unless specified otherwise.

MySpace RDF service - Yves

Yves — Thu, 13 Mar 2008 11:43:21 +0100

Hi Danny!

Indeed (I think Tom Morris made a similar comment at the Semantic Camp) how should we deal with "avatars"? For example, how would we make the connection between David Bowie and Ziggy Stardust?

Cheers!
y