ECRM OWL2 Draft
Georg Hohmann
recently we discussed (again) if we should lift the ECRM to OWL2. To
bring this discussion a step further, I created a first draft of the
ECRM in OWL2. You can find this version (temporarily) here:
http://erlangen-crm.org/110505/ecrm_110505.owl
There is no specific sublanguage called OWL2 DL. Instead, decidability
is guaranteed by using OWL2 Full with some restrictions (http://
www.w3.org/TR/2009/REC-owl2-syntax-20091027/#Ontologies). OWL2 (DL)
has the expressivity of SROIQ and comes with three profiles: EL, QL
and RL (http://www.w3.org/TR/owl2-profiles/). EL and QL do not support
all necessary features. For example, EL does not support cardinality
restrictions, QL does not support transitive properties. Therefore RL
would be the way to go, although there is now need to explicitly
define a profile we use as long as we consider the restrictions to be
DL compliant. The draft version of the ECRM OWL2 currently has the DL
expressivity SHIN(D).
Using OWL2 has some real benefits:
- We finally can use Protege4 for development which comes with a nicer
GUI and cleaner XML code.
- We can implement the so-called shortcuts of the CRM explicitly.
- We gain more expressivity.
- We have to maintain only one version (no "current" anymore).
The last point is feasible because of the new versioning syntax of
OWL2. The draft version is already built that way. Take a look at the
document header and the following page for more information:
http://www.w3.org/TR/2009/REC-owl2-syntax-20091027/#Ontology_IRI_and_Version_IRI
Comments?
Regards,
Georg
Thomas Bauereiß
I agree that OWL 2 has some real benefits compared to OWL 1, and that
OWL 2 RL is an interesting profile, as reasoning for OWL 2 RL can be
efficiently implemented using a rule-based system. I'd just like to
give some comments about OWL 2 RL.
First, like the other profiles, it also imposes several restrictions
on the expressions that can be used. In superclass expressions, for
example, no minimum cardinality restrictions (only 0/1 maximum
cardinality) and no existential quantification to a class expression
(i.e., someValuesFrom) can be used [1]. Both of these are used in the
current ECRM draft, as shown by the OWL 2 Validator [2], so if you
want to make ECRM compliant to OWL 2 RL (and can live without these
features), then some changes to ECRM will be necessary.
Second, if I understand the OWL 2 documentation correctly, there are
some differences in semantics between OWL 2 RL and OWL 2 DL.
_Syntactically_, OWL 2 RL is defined as a subset of OWL 2 DL [3], but
the rule-based reasoning uses the OWL 2 RDF-Based Semantics [4], while
OWL 2 DL is defined in terms of the OWL 2 Direct Semantics (see [5]
for a discussion of the differences and relationship between the two
semantics). I suppose that, in practice, this is not much of a
problem, as I think [6] states that under certain conditions, the two
semantics are equivalent, but it might be something to keep in mind.
But, as you said, it might not be necessary to select an OWL 2 profile
now. If the performance of generic OWL 2 DL reasoners is good enough,
then you can just go for OWL 2 DL, which will give you the benefits
you mentioned with tractable reasoning.
Best regards,
Thomas
[1] http://www.w3.org/TR/owl2-profiles/#Feature_Overview_3
[2] http://owl.cs.manchester.ac.uk/validator/
[3] http://www.w3.org/TR/2009/REC-owl2-conformance-20091027/#Document_Conformance
[4] http://www.w3.org/TR/2009/REC-owl2-profiles-20091027/#Reasoning_in_OWL_2_RL_and_RDF_Graphs_using_Rules
[5] http://www.w3.org/TR/owl2-rdf-based-semantics/#Appendix:_Relationship_to_the_Direct_Semantics_.28Informative.29
[6] http://www.w3.org/TR/2009/REC-owl2-profiles-20091027/#Theorem-PR1
On 11 Apr., 14:37, Georg Hohmann <ge...@hohmann.io> wrote:
> Dear all,
>
> recently we discussed (again) if we should lift the ECRM to OWL2. To
> bring this discussion a step further, I created a first draft of the
> ECRM in OWL2. You can find this version (temporarily) here:http://erlangen-crm.org/110505/ecrm_110505.owl
>
> There is no specific sublanguage called OWL2 DL. Instead, decidability
> and RL (http://www.w3.org/TR/owl2-profiles/). EL and QL do not support
> all necessary features. For example, EL does not support cardinality
> restrictions, QL does not support transitive properties. Therefore RL
> would be the way to go, although there is now need to explicitly
> define a profile we use as long as we consider the restrictions to be
> DL compliant. The draft version of the ECRM OWL2 currently has the DL
> expressivity SHIN(D).
>
> Using OWL2 has some real benefits:
> - We finally can use Protege4 for development which comes with a nicer
> GUI and cleaner XML code.
> - We can implement the so-called shortcuts of the CRM explicitly.
> - We gain more expressivity.
> - We have to maintain only one version (no "current" anymore).
>
> The last point is feasible because of the new versioning syntax of
> OWL2. The draft version is already built that way. Take a look at the
>
> Comments?
>
> Regards,
> Georg
Georg Hohmann
> First, like the other profiles, it also imposes several restrictions
> on the expressions that can be used. In superclass expressions, for
> example, no minimum cardinality restrictions (only 0/1 maximum
> cardinality) and no existential quantification to a class expression
> (i.e., someValuesFrom) can be used [1]. Both of these are used in the
> current ECRM draft, as shown by the OWL 2 Validator [2], so if you
> want to make ECRM compliant to OWL 2 RL (and can live without these
> features), then some changes to ECRM will be necessary.
That is a good point.
I didn't see that RL has such restrictions. Changing the ECRM this way
to be RL-compliant is not an option (in my opinion). So I think we
should stick with OWL2 (DL) as long as we feel a real need to go with
RL.
Regards,
Georg
Mark Fichtner
as a preliminary work for the OWL 2 version of the CRM we analyzed the shortcuts in the CRM. We found 18 shortcuts while 15 of them are more or less explicitly mentioned in the scope notes:
1) "P2 has type" is a shortcut for E1 -> P41i -> E17 Type Assignment -> P42 -> E55
2) "P43 has dimension" is a shortcut for E70 -> P39 -> E16 Measurement-> P40 -> E54 (according to the scope note of P43)
3) "P44 has condition" is a shortcut for E18 -> P34 -> E14 Condition Assessment -> P35 -> E3
4) "P1 is identified by" is a shortcut for E1 -> P140i -> E15 Identifier Assignment -> P37 -> E42 (according to the scope note of P1)
5) "P49 has former or current keeper" is a shortcut for E18 -> P30 -> E10 Transfer of Custody -> P28/P29 -> E39 (according to the scope note of P49)
6) "P50 has current keeper" is a shortcut for E18 -> P30 -> E10 Transfer of Custody -> P29 -> E39 (according to the scope note of P50)
7) "P51 has former or current owner" is a shortcut for E18 -> P24 -> E8 Acquisition -> P23/P22 -> E39 (according to the scope note of P51)
8) "P52 has current owner" is a shortcut for E18 -> P24 -> E8 Acquisition -> P22 -> E39 (according to the scope note of P52)
9) "P53 has former or current location" is a shortcut for E18 -> P25 -> E9 Move -> P26/P27 -> E53 (according to the scope note of P53)
10) "P55 has current location" is a shortcut for E18 -> P28 -> E9 Move -> P27 -> E53 (according to the scope note of P55)
11) "P56 bears feature" is a shortcut for E19 -> P59 -> E53 Place -> P53 -> E26 (according to the scope note of P56)
12) "P59 has section" is a shortcut for E18 -> P58 -> E46 Section Definition -> P87 -> E53 (according to the scope note of P58)
13) "P62 depicts" is a shortcut for E24 -> P65 -> E36 Visual Item -> P138 -> E1 (according to the scope note of P62)
14) "P105 right held by" is a shortcut for E72 -> P104 -> E30 Right -> P75 -> E39 (according to the scope Note of P105)
15) "P107 has current or former member" is a shortcut for E74 -> P144 -> E85 Joining -> P143 -> E39 (according to the scope note of P107)
16) "P109 has current or former creator" is a shortcut for E78 -> P147 -> E87 Curation Activity -> P14 -> E39 (guessed from scope note P109)
17) "P130 shows features of" is a dynamic asymmetric shortcut for E70 -> P15 -> E12 Production/E65 Creation -> P108/P94 -> E70 (guessed from scope note P130)
18) Everything else is a shortcut for E1 -> P140i -> E13 Attribute Assignment -> P141 -> E1
In some cases the shortcuts have subproperties. Thus they also have to be shortcuts because they inherit the more elaborated path from their superproperty.
19) As "P2 has type" is a shortcut, "P137 exemplifies" is a shortcut, too.
20) As "P1 is identified by" is a shortcut, "P102 has title", "P131 is identified by", "P48 has preferred identifier", "P78 is identified by" and "P87 is identified by" must be shortcuts
21) As "P105 right held by" is a superproperty for "P52 has current owner", therefore P52 is a shortcut for both more elaborated paths.
22) "P130 shows features of" is a superproperty for "P73 has translation" and therefore P73 is a shortcut, too.
Now some questions for all of you:
- The scope note of P1 has a contradiction. The range of "P1 is identified by" is E41. The scope note tells that "P1 is identified by" is a shortcut for E1 -> P140i -> E15 Identifier Assignment -> P37 -> E42. The end of this more elaborated path is E42 and not E41 (range of P1). So either the range of P1 is wrong or the range of this P37 is wrong or the definition of this more elaborated path is wrong. What do you think?
- "P45 consists of" and "P72 has language" seem to be some kind of shortcut, too. Their ranges (E57 Material and E56 Language) are subclasses of E55 Type, therefore 1) can be used. However the semantic of this more elaborated path is questionable in this case.
- Have we missed any shortcuts?
It would be great to see someone make use of these shortcut-rules.
Regards,
Mark
-------- Original-Nachricht --------
> Datum: Mon, 18 Apr 2011 11:27:03 +0200
> Von: Georg Hohmann <ge...@hohmann.io>
> An: erlang...@googlegroups.com
> Betreff: Re: [ECRM] Re: ECRM OWL2 Draft
> --
> You received this message because you are subscribed to the Google Groups
> "Erlangen CRM" group.
> To post to this group, send an email to erlang...@googlegroups.com.
> To unsubscribe from this group, send email to
> erlangen-crm...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/erlangen-crm?hl=en-GB.
>
--
Empfehlen Sie GMX DSL Ihren Freunden und Bekannten und wir
belohnen Sie mit bis zu 50,- Euro! https://freundschaftswerbung.gmx.de