- From: Dan Brickley <Daniel.Brickley@bristol.ac.uk>
- Date: Sun, 12 Dec 1999 18:44:04 +0000 (GMT)
- To: Guha <guha@epinions-inc.com>
- cc: Dan Brickley <danbri@w3.org>, RDF Interest Group <www-rdf-interest@w3.org>, swick@w3.org
On Sun, 12 Dec 1999, Guha wrote: > The reified triple is not meant in anyway to assert > that the triple is part of any graph. So, for example, > one could state : > > believesIn(Fred, shape(Earth, Flat)) > > without asserting that the earth was flat. Sure, that's the entire point of it, right? Web of Trust and all that... I guess I wasn't being clear enough. I'll try again: A reified statement such as that above... call it [s1]... [s1] - type -> rdf:Statement [s1] - subject -> [Earth] [s1] - predicate -> [shape] [s1] - object - > Flat] Informally, [s1] represents the assertion that the earth is flat. From talking to a fair number of RDF implementors, there appears to be two alternative readings of what's going on here. Everyone agrees that [s1] is the statement that the world is flat, but there are some who might expect an RDF store to represent distinct statement resources [s1], [s2], [s3] for Dan, Guha and Ralph's respective assertions that the earth is flat. On this picture, each statement ('assertion','claim') is made by some source, eg. [s1]--assertedBy--> [http://epinions.com/people/guha] [s1]--pics2:on--> [2000-04-01] While [s2] and [s3], although sharing the same subject/predicate/object outgoing arcs, would have distinct identity within the Web and different URIs. This view feels natural in particular to those who are implementing RDF triple stores as rows in relational dbase tables, since a unique ID on each triple/row can be used to hook up to the source/attribution of that piece of knowledge. (eg. we could in principle define an RDBMS view on the subset of triples that were asserted by Guha on 2000-04-01). The alternative view is that the ideally is just one Web resource representing the abstract assertable claim [s1]. My Question (3) below asks whether, if this is the case, we could explore canonicalisation of the URIs for these to aid aggregation. Recap... First reading: statements with same subject/predicate/object occur many times, for the various belief and disbelief scenarios in which they occur[*] Second reading: identity condition for statements is having the same subject/predicate/object. There's really only one statement resource for each state-able triple -- if we use a variety of URIs instead of one that's a pragmatic hack rather than recognition that there are multiple statements. Dan [*] I see my earlier mistake now: I gave impression by using term 'statings' that 'believes' was the only relation an agent might have to a statement. We could of course enumerate a whole suite of attitudes that agents might have to propositions: believes, feares, denies, suspects... > Dan Brickley wrote: > > > Here's a puzzle: is an RDF Statement (ie. a triple in reified form, an > > instance of the class Statement) a representation of some "stating" of a > > subject/predicate/object triple, or a representation of the abstract > > statement that is being asserted? > > > > I believe the model spec is a little unclear on this, but that the answer > > should be that there are two distinct concepts ('assertions events' or > > 'statings' versus 'statements') and that our reified 'RDF Statement' > > objects model the latter. A particular statement, then, could be asserted > > by different people on different dates in different contexts. Each of > > these claims might have PICS-labelesque attributes such as being 'by' > > some agent, 'on' some date. Whereas the statements themselves are > > timeless. > > > > Three questions: > > > > 1. Does this distinction seem clear? > > > > 2. Does my reading of the meaning of 'RDF Statement' seem correct? > > > > 3. If so, can we use algorithms such as those Sergey proposes to assign > > globally unique, identical identifiers to each statement to facilitiate > > data aggregation? (eg. any occurance of [bill clinton] > > --livesIn->[America] could have a canonical statement URI generated, > > like uuid:423423532453443 such that all graphs including reference to this > > statement could usefully be aggregated). Each 'stating' of the claim that > > [bill clinton]--livesIn->[America], by contrast, could have a different > > URI since it would happen of a different date by a different agency. > > > > Am I making any sense? > > > > Dan > > > > -- > > danbri@w3.org > >
Received on Sunday, 12 December 1999 13:46:48 UTC