Re: struggling with ASK

On 11/30/11 5:02 AM, Dan Brickley wrote:
> On 30 November 2011 02:24, Kingsley Idehen<kidehen@openlinksw.com>  wrote:
>> On 11/29/11 6:03 PM, Peter Williams wrote:
>>
>> Thanks folks. I think Im nearly done.
>>
>> Love now to see someone's (*off topic*) ideas for walking the foaf:knows
>> link chain, using the Euler reasoner.
>>
>> Re. walking the chain via SPARQL albeit Virtuoso SPARQL specific, see:
>>
>> 1. http://lod.openlinksw.com/b3s/search.vsp?q=6
> Nice technically! How much of this is do-able in SPARQL 1.1 without
> extensions btw?

Pulling that off still isn't part of SPARQL hence the extension. Note, 
my instincts are always to use the standards first before resorting to 
extensions. The link above is more than 4 years old.
>
> One not-exactly-technical concern: we should be careful not to
> overload the meaning of 'foaf:knows'.

Naturally, this pattern isn't foaf:knows specific.

>   It is by design weak, an
> entry-level social link.

Yes.

>   It is not supposed to be a risky or
> emotionally draining thing to decide to assert it.

Yes.

>   No "am I really
> friends with this person" teen-blogger-angst. Taken in aggregate (as
> we do here) a big pile of foaf:knows assertions about some party might
> carry more information, but the basic idea is just that some person X
> knows some person Y, in the weak sense of 'have had some reciprocated
> interaction' with them.

When WebID apps start to propagate befriending and reciprocity will have 
better fidelity for these kinds of graph analytics. Once this starts, 
we'll add addition social network oriented examples.


> It might be read as 'asserts common humanity
> of', except in practice people don't do much checking, and accepts
> bots and scripts and spammers a bit too freely.
>
> In practice as exported from various social Web sites, it means roughly
>
> * party X has 'added' party Y in the site
> * party Y has glanced quickly and the name and photo details the site
> showed for X
> * party Y confirms to the site that they know/accept/allow X to be
> linked to them (as a 'friend' or some such increasingly meaningless
> label)

Yes, clear about all of that.

>
> the slow rise of friend lists, groups, circles etc will shake things
> up a bit. But nothing yet strongly implies trust, or any careful
> checking of evidence.
>
> I do think there's something that could be done in which people more
> explicitly assert that they've seen evidence supporting some binding
> of real-life identity (nym'd or not) with online accounts. So for
> example if skype:hs12341234 and I have a video chat, and I am
> persuaded that the other party is Henry, and not under obvious duress,
> and he says 'yes this is my personal skype account and nobody else
> controls it', ... I'd like some idiom in RDF to log and possibly share
> that information so that others might choose to partially rely on it.

As per earlier comment reciprocal links will help. Ultimately, people 
will leverage associated ontologies that deal with these matters.

>
> I have a vague feeling that's where the foaf:knows transitivity theme
> is expected to go; however the meaning of that relationship is weak
> both by design and by common implementation, that I'm not sure it'll
> carry the weight of expectations that are being heaped upon it.

My link is more about demonstrating what's possible :-) Ultimately, 
people will perform graph analysis using appropriate ontologies. 
Examples that veer in this direction include:

1. http://smiy.sourceforge.net/wi/spec/weightedinterests.html -- 
weighted interests ontology
2. http://smiy.sourceforge.net/cco/spec/cognitivecharacteristics.html -- 
cognitive characteristics ontology .


Kingsley
>
> cheers,
>
> Dan
>
>


-- 

Regards,

Kingsley Idehen 
Founder&  CEO
OpenLink Software
Company Web: http://www.openlinksw.com
Personal Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca handle: @kidehen
Google+ Profile: https://plus.google.com/112399767740508618350/about
LinkedIn Profile: http://www.linkedin.com/in/kidehen