On 11/14/12 4:16 PM, Ed Barkmeyer wrote:
> Kingsley Idehen wrote:
>> Ed and other respondents,
>> I prefer to look at the entity relationship semantics on the Web as
>> being akin to screen resolution fidelity. You have high and low
>> resolution. There are data spaces within the Linked Open Data cloud
>> where the fidelity or entity relationship semantics are very high and
>> discernible to humans and machines. Of course, there are enclaves where
>> the aforementioned semantic fidelity is very low. Such is the nature of
>> the Web.
>>
> The first question is whether and how the Web technologies in use
> support the identification of the 'entity relationship semantics' of a
> link. The meaning of the link in Pyotr Nowara's email today was not
> clear to some of the participants on this exploder, even though it was
> embedded in English text that was presumably capable of expressing its
> exact relationship to the subject of the email. (01)
Yes they do. (02)
You make the very point I make repeatedly when someone shows me a chunk
of XML or JSON and then verbally explains to me the entity relationship
semantics. For instance, last night, someone was verbally speaking about
the relationship between 'Stevie Wonder' and a musical instrument. My
response was: if you have the entity relationship semantics done right,
I would discern that from the chunk of JSON myself i.e., no semantic loss. (03)
> By comparison, formal
> links in HTML documents are nothing more than short strings with
> associated URIs. What is the relationship between the hyperlink text
> and the information body at the URI? The problem in such links is not
> whether the linked information is faithful to the intended E-R
> semantics; the problem is in conveying the intended E-R semantics at all. (04)
Yes, not disputing that. (05)
>
> In a language like RDF, the stated semantics of the use of a URI term is
> that the term designates some concept and the URI can be dereferenced to
> some resource that facilitates our understanding of that concept. (06)
Yes. My very response to the person presenting my with his JSON blob
about 'Stevie Wonder' . (07)
> There
> one can talk about fidelity, because the intrinsic semantics of the link
> is well-defined. That is a major difference from an HTML hyperlink or a
> URI value of an XML attribute. (08)
Yes. (09)
>
>> The Web's ultimately advantage is that crowd-sourcing is intrinsic. The
>> number of subject matter experts contributing to the Linked Data Cloud
>> is growing rapidly too :-)
>>
> "Crowd sourcing a la Wikipedia" and "Crowd sourcing a la Google" are two
> entirely different things. (010)
Yes, and I mean something better than both. (011)
> The committed experts contributing to
> Wikipedia articles work actively to suppress the committed ignoramuses.
> That doesn't prevent Wikipedia articles from being inaccurate or
> debatable or one-sided, because there are many fields in which that kind
> of information variance is typical of "accepted truth" in the trade. It
> does prevent clear misinformation from surviving very long. By
> comparison, Google's approach is essentially a political contest -- the
> most commonly referenced links are presented first, although Google is
> also doing some kind of credibility control. This kind of crowd
> sourcing returns what most people accept, rather than what experts
> accept. And it is ripe for exploitation, as Doug Foxvog pointed out:
> "Never underestimate the power of stupid people in large groups." (012)
Yes. (013)
>
> My experience in standards development has significant parallels. There
> are development groups led by effective organizers with expert
> contributors who work together to produce something of value. There are
> development groups led by effective organizers deliberately ignorant of
> other work who build groups with motivation but little relevant
> expertise to produce would-be standards of no real value. And there are
> groups consisting of a few real experts and a few uneducated leaders and
> a body of lesser folk who are regularly asked to choose between the
> right way and the wrong way without having the background to be capable
> of distinguishing them. The webby standards bodies themselves fall into
> all three categories. So we have every reason to suppose that the
> highly linked Web will have all of those characteristics. (014)
Yes, no doubt. I concur from my own personal experience with a number of
W3C groups. (015)
>
> It seems to me (and I think this is what TBL had in mind in the treatise
> John Sowa cited) that the experts can be expected to find and recognize
> other expert work, and thus their links will be reliable and as
> well-defined as the technologies permit. So, good information will link
> to good information, for the most part. (016)
Yes, over time, since the density of the Web ultimately increasing the
potential for serendipitous discovery. (017)
> Bad information linked to good
> information, and bad information linked to bad information, and good
> information badly linked will be much more common, because expertise is
> a much rarer commodity than "information" on the Web. (018)
Yes, but it is going to be easier to spot and quarantine bad data. This
is why we use pragma's in our SPARQL implementation to make inference
context conditional. Can't just apply owl:sameAs semantics because one
stumbles across those kinds of claims in the LOD cloud, for instance. (019)
> The problem will
> always be finding the good starting points. And, like the literature of
> certain trades, there will unfortunately be brilliant insightful
> contributions that are inadvertently suppressed because they are not
> mainstream and get few citations, and conversely, there will be
> "revolutionary" treatises that get lots of citations by being
> provocative rather than accurate. (020)
Yes. (021)
> "Even bad publicity is good", if what
> you are counting is site visits, not content value. (022)
Yes. (023)
>
> So I do not see the crowd-sourced linked information space being a
> significant improvement over the current situation. (024)
It is, because Linked Data isn't a read-only proposition. In addition,
inference context itself is controllable as per my earlier comments. (025)
> The Linked Open
> Data technologies just make the linking easier; they don't improve the
> capture of the semantics of the links or the quality of the linked
> resources. (026)
They give you something to accept or fix. When you fix, it can occur in
your own data space. Basically, you can agree to disagree, without
introducing inertia. (027)
> There will be valuable entities clearly related to other
> valuable entities; valuable entities somehow related to other valuable
> entities; and a large body of work in which one or the other entity is
> mostly garbage and the quality of the link is irrelevant. And that is
> where we are now. (028)
Sorta. (029)
> What we need is mechanisms for creating high-quality
> resources, mechanisms for recognizing the quality of resources, and
> mechanisms for improving the expressiveness of the links. (030)
We do have that now.
>
> As near as I can tell, the current web-think is that RDF and its
> spinoffs will somehow do this for us (after only 14 years). (031)
Yes. (032)
The 14 year odyssey is a product of the many issues John raises in his
critique of the poor choices of yore. Things are much better now, and
improving more often than not. (033)
>
> The alternative being promulgated by several standards bodies takes the
> general form of an "architecture", or a "framework" or a "best practice
> guideline", for what gets linked to what and how, and one can better
> understand the relationships expressed by the links if one understands
> the architecture/practice being used. This approach is great for the
> "in-crowd" who are part of the development activity, but the resulting
> artifacts are less clear to the web tourist who enters one of these
> well-developed structures in the middle. (034)
Yes. (035)
> Some link in the resource or a
> related resource points to the architecture, and if you accidentally
> find it, you may be able to determine how you should have followed the
> links.
>
> -Ed
> (036)
Some links to illustrate what exists in relation to my view points: (037)
1.
http://linkeddata.uriburner.com/about/id/entity/http/www.linkedin.com/in/kidehen (038)
-- Linked Data URI that denotes me in a profile graph derived from
LinkedIn (you can access this entity relationship graph in a variety of
formats via cURL or via the links in the page footer) ; the TBox links
also resolve to descriptions (039)
2.
http://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Flinkeddata.uriburner.com%2Fabout%2Fid%2Fentity%2Fhttp%2Fwww.linkedin.com%2Fin%2Fkidehen&gp=18&go= (040)
-- Faceted browsing and deeper follow-your-nose oriented profile page (041)
3.
http://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Flinkeddata.uriburner.com%2Fabout%2Fid%2Fentity%2Fhttp%2Fwww.linkedin.com%2Fin%2Fkidehen&gp=20&go= (042)
-- specific page (043)
4.
http://linkeddata.uriburner.com/describe/?url=http%3A%2F%2Fwww.openlinksw.com%2Fschemas%2Fcert%23hasCertificate (044)
-- shows you the 'Owns' property/predicate definition and you can
drill-down deeper to our security token ontology which links to other
ontologies (045)
5. http://bit.ly/TH4Mhf -- examples of subject matter experts from the
Pharma industry exploiting Linked Data (046)
6. http://slidesha.re/UGg18k -- presentation I gave a last nights
Semantic Web Meetup re. Linked Data, RDF, and a Semantic Web (basically
explaining why conflation always hurts the conflated) . (047)
-- (048)
Regards, (049)
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 (050)
smime.p7s
Description: S/MIME Cryptographic Signature
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J (01)
|