Re: [UCR] Rich KR Use Case

Jim,

OK, I see what you are getting at - the way the example is presented 
could be interpreted as suggesting that all variables in rules must be 
bound to named individuals (a condition sometimes referred to as "DL 
safety"). This is a good point. I agree that it would also be better to 
use a more directly web related example, such as one coming from the 
web services domain, and make sure that it explicates the differences 
between the two cases (with/without a DL safety condition) and the need 
for semantic compatibility. It should also be made clearer that the use 
case shows the need for rules *and* ontology languages to be used 
together to achieve what neither can (typically) achieve alone - 
bringing us back again to the need for semantic compatibility.

I will try to put together a new example along the lines I have 
suggested above, but I won't be able to distribute it or update the UCR 
doc for a couple of days (I am travelling all day tomorrow). Regarding 
the syntax, I will stick to abstract syntax (if I try to generate RDF 
it will be full of bugs), but I will have no objection if you or some 
other kind person wants to add the corresponding RDF syntax.

Ian



On 22 Feb 2006, at 23:17, Jim Hendler wrote:

> Ian - thanks for your comments - I mostly agree.  My comments inlined 
> below (I'll use JAH as a tag so they're easier to find in the long 
> thread)
>
> At 21:18 +0000 2/22/06, Ian Horrocks wrote:
>> On 20 Feb 2006, at 17:58, Jim Hendler wrote:
>>
>>>
>>>  I objected at the beginning to the way this was going, I object now 
>>> to how it sounds.  The problems are that (1) this implies that Rich 
>>> KR is somehow better and more valuable on the web (largely 
>>> unproven),
>>
>> Jim,
>>
>> The set of use cases lumped together under this title are pretty 
>> heterogeneous. Some do, I agree, emphasise KR on the web more than KR 
>> for the web, but I have been assured that the success of this kind of 
>> application constitutes a success for the semantic web as a whole, 
>> and so could be seen as a relevant input for RIF (not to mention the 
>> fact that publishing/sharing ontolgies on the web is an important 
>> aspect of many of these use cases). Of the examples available on the 
>> Wiki, only the "Uncle" example has been chosen (not by me) for 
>> inclusion in the UCR document. I guess that this is because it is 
>> well known and relatively easily understood (although I note your 
>> comments below). Similar examples also arise in the web services 
>> domain (amongst others), and it would be a good idea to at least 
>> refer to these in order to strengthen the motivation in general and 
>> the "for the web" motivation in particular.
>>
>
> JAH: My fear was that the uncle example is more "misunderstood" than 
> most others, and thus people may not read it as clearly - one of the 
> others would have made me happier.
>
>>
>>>  (2) it takes an unflattering approach to describing OWL in its 
>>> emphasis,
>>
>> This didn't really strike we when I first read it, but you could be 
>> right. It would certainly be a good idea to emphasise the fact that 
>> OWL can also express things that are not typically expressible in 
>> rule languages, i.e., that the two paradigms have complementary 
>> expressiveness. The sentence "A typical example, as detailed below, 
>> is the use of rules to describe complex relationships between binary 
>> predicates which cannot be captured in ontology languages alone." 
>> should also be fixed - it may be true of OWL, but it isn't true of 
>> ontology languages in general.
>>
>
>
> JAH: yes.
>
>>
>>>   and (3) it doesn't emphasize the important point it is trying to 
>>> make (that sometimes extending OWL with rules is a good thing, but 
>>> the rules syntax needs to be commensurate with the syntax OWL is in,
>>
>> For my taste the *really* important point (which also isn't 
>> emphasised) is the need for semantic compatibility with RDF and OWL. 
>> I'm not sure that I understand what is meant by "commensurate with 
>> the syntax OWL is in", and if it means what I think it means (i.e., 
>> RDF/XML syntax), then I'm not sure that I agree.
>>
>
> JAH: Yes, I would prefer same syntax (so I can have a file which mixes 
> my rules and my OWL together when I define both using SWOOP or some 
> such editor) but the point I was after here is more the semantic 
> compatibility.
>
>>
>>>  (4) The syntax used in the example is not the normative exchange 
>>> syntax for OWL  and this further obfuscates the point being made.
>>
>> Again, I guess that you are asking for the example to be given in 
>> RDF/XML syntax. The trouble with this is that human readability would 
>> then be very poor, even for those reasonably familiar with RDF (which 
>> many in the WG and beyond may not be). I guess that the abstract 
>> syntax was chosen in an effort to *avoid* obfuscation (I'm not sure 
>> about this as I wasn't the originator of this example).
>>
>
> JAH: someone who reads the OWL guide and Reference documents should be 
> able to read the OWL in this document - I'd rather see compatibility 
> than terseness if the goal is to get people to consider using these 
> languages together.
>
>>>
>>>  I do not object to having a section (Renamed "extending OWL") which
>>
>> I'm not very keen on the Rich KR name either and, speaking for 
>> myself, would be happy to see it changed to "extending OWL".
>>
>>>   1 - shows how rules can extend something in OWL (I think the uncle 
>>> example is very misleading, by the way, unless you want to get into 
>>> a discussion of safe grounding - seems odd to have to insist on 
>>> including the "person(?y) and person(?x) in the rules when the WG 
>>> hasn't even addressed this issue)
>>
>> I agree that "person(?y) and person(?x)" may not be necessary, but it 
>> doesn't seem to be hurting - surely it is reasonable to expect the 
>> sibling and child in question to be Persons given that an Uncle must 
>> be a Man?
>
> JAH: Ian, if I understand the rules literature as well as I think I 
> do, these aren't there to be "typing" statements, they're there to be, 
> for lack of a better word, "grounding" statements - i.e. many rule 
> languages won't work without them  - even if I say /sibling/ has the 
> domain and range of /person/, I still need the person(?x) to make sure 
> the ?x is found and bound in most polynomial rule languages (and maybe 
> all safe ones) - I'm still grappling with the exact technical 
> vocabulary for this, but my understanding is that some people seeing 
> this would think we meant something we weren't attempting to convey.
>
>>
>>>   2 - that does not have any italics nor use the incorrect word 
>>> impossible  (btw, it is more than possible to do this in OWL,, where 
>>> I can extend syntax, it is not possible to do this in OWL DL as 
>>> written)
>>
>> Well, I am as big a fan of OWL as the next person, but I do believe 
>> that it is "impossible to describe the desired relationship between 
>> the hasSibling, hasChild and hasUncle properties in OWL" (either DL 
>> or Full). The desired relationship is such that the uncle 
>> relationship is entailed for any pair of individuals in the relevant 
>> "indirect" relationship. Of course OWL could be extended to include 
>> such expressive power (and in fact efforts are already underway to 
>> define an extension to OWL DL that includes a limited form of role 
>> composition that does not make the language undecidable) - perhaps it 
>> would be worth mentioning/discussing this in the text.
>>
>
> JAH: the point I was after was the previous one you mentioned about 
> them having different coverages - we don't say anything about the 
> stuff in OWL being <em>impossible</em> in RIF, and thus it just seems 
> like a point was being made that it wasn't necessary to make
>
>>>   3 - makes the last sentence, which is more important than all the 
>>> rest of the example, more evident
>>
>> I agree that this important point is not very prominent (maybe it was 
>> deemed to be implicit given the context).
>>
>>>   4 - is retitled to emphasize that this is about the standard 
>>> format for extending, not about the extending per se (there are many 
>>> other ways to consider extending languages w/or w/o rules)
>>
>> Again, I would say that the crucial point is semantic compatibility. 
>> I agree that there are many other ways in which OWL could be 
>> extended, and that this could/should be mentioned in the text. It 
>> doesn't seem unreasonable to focus on rule extensions in this case, 
>> however.
>>
>
> JAH: point I was making is the one you made earlier, that the focus 
> should be on compatibility, not KR.
>
>>>
>>>  btw, less anyone wishes to claim my objections here are more 
>>> political than technical, I confess you are absolutely right.  THe 
>>> WG process is inherently political, and that is why I pay to belong 
>>> to the W3C so I can help get the politics right.
>>
>> I hope I focussed on the technical objections - I prefer to leave 
>> political to the experts :-)
>
> JAH: I think I've been insulted very subtly - nicely done Ian!
>
>>
>> One point that I remember well from our WOWG days is that, in 
>> response to criticism of proposed WG texts, you would usually ask the 
>> critic to suggest alternative wording. Perhaps you would like to do 
>> so in this case?
>>
>
> JAH: umm, well, actually my preference would be to delete this use 
> case.  However, since many in the WG seem to want it, I still hope 
> there is a consensus space where I would be willing to abstain. but 
> until then, my proposed rewording is "delete this section" - sorry.
>
>> Regards,
>>
>> Ian
>>
>>
>>>
>>>   -JH
>>>
>>>
>>>  -- Professor James Hendler			  Director
>>>  Joint Institute for Knowledge Discovery	  	  301-405-2696
>>>  UMIACS, Univ of Maryland			  301-314-9734 (Fax)
>>>  College Park, MD 20742 http://www.cs.umd.edu/~hendler
>>>  Web Log: http://www.mindswap.org/blog/author/hendler
>>>
>
> -- 
> Professor James Hendler			  Director
> Joint Institute for Knowledge Discovery	  	  301-405-2696
> UMIACS, Univ of Maryland			  301-314-9734 (Fax)
> College Park, MD 20742	 		  http://www.cs.umd.edu/~hendler
> Web Log: http://www.mindswap.org/blog/author/hendler

Received on Friday, 24 February 2006 20:45:14 UTC