- From: Kingsley Idehen <kidehen@openlinksw.com>
- Date: Mon, 18 May 2009 11:45:53 -0400
- To: David Huynh <dfhuynh@alum.mit.edu>
- CC: John Graybeal <graybeal@mbari.org>, Linked Data community <public-lod@w3.org>, semantic-web@w3.org
David Huynh wrote: > John Graybeal wrote: >> I think a key point of David's original post, which I would like to >> emphasize, is that if I use a URL to refer to a web page, the owner >> is generally either readily visible (in the URL) or discoverable (via >> domain lookup). When I represent a term as a URL, it is at least >> known who is serving that URL; and often the _developer_ of the term >> can be derived intuitively (e.g., by dereferencing the URL in a >> browser). These are positive social outcomes, and encourage further >> adoption of the term. >> >> When I represent a term as a URN or other URI that can not be >> dereferenced directly in a browser, almost all of that social context >> is lost. The responsible party can be found only by manually parsing >> the URN, going to an obscure (to most) web page, manually looking up >> the URN authority. The developer of the particular term may be >> discoverable from the rest of URN -- but any semantics embedded in >> the rest of the URN, if any, can only be known by finding and reading >> the materials from the application of the responsible party. And the >> actual metadata for the term can only be found by discovering, >> through close analysis of the application or some other way, a 'magic >> lookup URL' so a browser can look up the URN and provide additional >> information about it. >> It isn't that these problems can't be solved; eventually global URI >> resolution will probably be available with browsers thanks to tricky >> and consensual underlying technologies. But the initial >> specifications as rolled out provided no standard way to solve them, >> so until the marketplace converges, the social conventions available >> with URLs are not supported. >> >> I would quibble with some of the details of David's original >> argument. Someone other than Company X can have the companyx.nnn >> domain, for example. And semantic systems could use 'term rank' >> methods to derive the most important URI for a term like 'microsoft', >> thereby achieving at least one of Google's neat tricks in the >> semantic realm. But overall, the adoption of semantic and other >> URI-based technologies in the human world will inevitably lag that of >> http URLs, until the technologies are so fully developed that all >> these 'weaknesses' (strengths in other respects, of course) are fully >> addressed. >> > Well put, John! Thank you--that's what I meant to say. You and I are > on the same page. > > David > > > David / John, Still don't get how you don't see that Entity Rank is about what you are saying. You can only use link coefficients in the data space to determine what entity is referenced the most, and then combine that with associated literal objects that have text frequencies for a given text pattern en route to making the first result set (what we do). We expect the use to navigate by Type and/or Properties to find what they want. This is what we actually do in real life as humans. We never say: Find me Literals "Red Ball", we say and process: Find me an Object of Type ball with Color property: Red. Our UI isn't great, but we have covered the core of this matter, really. -- Regards, Kingsley Idehen Weblog: http://www.openlinksw.com/blog/~kidehen President & CEO OpenLink Software Web: http://www.openlinksw.com
Received on Monday, 18 May 2009 15:46:49 UTC