Re: Web Rule Language - WRL vs SWRL

On 30 Jun 2005, at 04:11, Michael Kifer wrote:

>
> Ian Horrocks wrote:
>>
>> On 29 Jun 2005, at 22:37, Michael Kifer wrote:
>>
>> [snip]
>>
>>>>>>> N3 is essentially a different syntax for F-logic and its 
>>>>>>> extensions
>>>>>>> (but
>>>>>>> N3's semantics is defined by use cases ;-). As far as I can tell,
>>>>>>> with
>>>>>>> each
>>>>>>> new presentation that I hear N3 is moving in the direction of LP.
>>>>>>
>>>>>> I think that we should stick to discussing how things actually 
>>>>>> *are*
>>>>>> rather than directions in which you hope/believe they might be
>>>>>> moving.
>>>>>
>>>>> I am discussing things as they already are. N3 now has a form of
>>>>> SNAF.
>>>>
>>>> Maybe, but this doesn't make it LP - we have long since known how to
>>>> support a form of SNAF in DL using the so-called K operator.
>>>
>>> What makes something an LP language?
>>
>> You tell me - you are the one who is claiming that things are "moving
>> in the direction of LP".
>
> Let me remind you that 5 lines above you said
> "this doesn't make it (N3) LP". So, by CWA (albeit not FOL) you must 
> have a
> definition of LP in mind.
> Now, 6 lines above the current line I asked you what is your 
> definition?
> Instead of answering you ask me back. We are having a hell of a
> conversation here.

I only provided one example of a language that I believe is *not* LP; 
this doesn't mean that I can define what LP *is*. You are the one who 
seems to have a set of sufficiency conditions, and claim that N3 is 
moving/has moved in a direction that will satisfy these conditions. If 
you say what these conditions are then we can determine if they are 
satisfied by a DL that includes the K operator.

>
>
>>>> So, an LP language would find an entailment that is *not* supported 
>>>> by
>>>> RDF semantics (which would allow for models in which John had other
>>>> children). Ergo, LP is semantically incompatible with RDF.
>>>
>>> You misunderstood my reply.
>>> I meant that I acknowledged your point about more expressive 
>>> languages
>>> not giving the same answers when they are both applied to the same
>>> dataset.
>>>
>>> RDF by itself doesn't have any queries defined over it, so it is
>>> meaningless to claim that a query language L over RDF syntax is
>>> incompatible with RDF.
>>
>> This is not what is being claimed. What is being claimed is that an 
>> RDF
>> triple of the form <x,P,y> is *not* semantically equivalent to the LP
>> rule P(x,y). This is the equivalence that is implicit in the layering
>> that we have been discussing (i.e., LP on top of DLP on top of RDF),
>> and which leads to LP entailments that are not valid w.r.t. the
>> semantics of RDF (or DLP).
>
> Yes, they are equivalent within the RDF subset.*
> Withing this subset, the only queries that you can ask is whether a 
> set of
> triples (bodyless rules) imply another triple (bodyless rule).

Sigh! For your convenience, I will repeat verbatim what was written in 
my email of 28th June:

> Regarding DLP under Horn (equivalently DL/FOL) and LP semantics, it is 
> true that *under certain assumptions* the two semantics are 
> indistinguishable, but this does *not* mean that they are equivalent. 
> The assumption being made in this case (as Bijan has pointed out) is 
> that only ground entailment will be considered. This is a rather 
> strong assumption: it is common to compensate for weak representation 
> languages (such as DLP) by providing much more expressive query 
> languages (e.g., the case of SQL), and with more expressive languages 
> we can easily distinguish two different semantics (as illustrated in 
> [1]).
>
> This is particularly important in a layered architecture, where the 
> increased expressive power of higher layers will naturally allow the 
> two different semantics to be distinguished. This was the main point 
> of [1], i.e., that we should not fool ourselves about the degree of 
> interoperability that would be provided by the architecture proposed 
> in the so-called "updated layer cake". Such interoperability only 
> exists so long as neither tower extends above the DLP level *and* so 
> long as we only allow for very weak query languages; DLP does not, 
> therefore, provide a suitable foundation for a layered architecture, 
> as higher layers (or more expressive query languages) might make false 
> assumptions about the intended semantics of DLP ontologies.

and your reply:

> This wasn't clear from the paper.
> In any case, the claimed interoperability doesn't extend to the more
> powerful languages.

As you know, the "updated layer cake" shows a single DLP layer. At 
best, this should be split into two separate languages:  DLP (with FO 
semantics as per the paper) and DLP-Datalog. Otherwise the layering 
breaks down, because treating DLP rules as LP rules is clearly not 
correct. In fact, with FO semantics (as per the paper), DLP rules *are* 
equivalent to OWL axioms which *are* equivalent to SWRL rules. 
Similarly, DLP-Datalog rules *are* equivalent to Datalog rules and so 
on. These equivalences (the two towers) can't *both* work for a 
*single* DLP language as is implied by the "updated layer cake".


>
> *Modulo the blank nodes. The post-facto RDF semantics treats blank 
> nodes as
>  head-existential, which is outside of LP. But there is another, 
> LP-style
>  semantics for blank nodes.

We seem to be back to discussing RDF with a different semantics than 
the one it *actually* has. If we assume that it would be possible to 
give RDF syntax an alternative LP style semantics, then we would have 
two completely separate language towers, one based on RDF and the other 
based on RDF-LP. This was *exactly* the point we were making in our 
paper.

Ian



>
>
> 	--michael
>
> PS. By saying that your point is well-taken, I acknowledged 
> consistency of
>     your argument about two language extensions being incompatible with
>     each other.
>     But I was talking about a different definition, that of one 
> language
>     being an extension of another. This is a standard definition and
>     according to it LP is an extension of DLP, and DLP is an extension 
> of
>     RDF (modulo the blanks).

Received on Thursday, 30 June 2005 10:59:33 UTC