Hi Steve and Cam,
I have a question re: your design of the current Semantic Darwin Core ontology. (And this is assuming that with the published 0.2 version I have the latest one in hand.)
My understanding is that all classes in DSW are imported from either DwC or DC (or FOAF), in order to reuse those terms. While that's a good idea in principle, it seems that DSW is actually not saying much new (in a semantic sense) about them, except to declare them to be OWL classes, and to assert them as disjoint from each other (or equivalent in one case). DSW then adds a variety of object (and some data) properties, which distinguish themselves from those in DwC by declaring domain and range axioms for them. But that doesn't say anything about the classes either, nor does it, I would argue, about the properties - domain and range constraint really only say something about the instances for which one asserts those properties.
So by itself the DSW won't allow me to infer anything about the classes and properties in the ontology (aside from disjointness), though it will allow me to make more inferences about instance data to which it is applied than DwC would. And those additional inferences would consist only of the instances' class memberships (and their non-memberships).
I'm wondering a) whether I'm missing something here and am in error, and if not, b) whether the above was indeed the extent of what you wanted to achieve with DSW. Either way, what are your current plans with the ontology? It doesn't seem to have changed for a while.
(And please forgive me if this isn't the right list to post to - I couldn't find a DSW-specific one on the Google code homepage.)
Cheers,
-hilmar
Without regard to your position---for which I'd have to look more carefully at DSW to agree or not---my own present position is that reasoning on data described by an ontology is way more interesting than reasoning on the ontology itself.Well, sometimes. For example, due to the strong typing they provide by putting domains on properties (of which I disapprove, but that's another story), you can infer types in the data whether they were avowed or not. In particular, they may have enough in play to detect avowed type errors in the data just because the reasoner would signal an inconsistency.
My current hobby horse is to evaluate ontologies with competency questions, preferably phrased formally with SPARQL or some other query language. When I make those, the data seems to be foremost on my mind.
Bob
On Wed, Jan 25, 2012 at 8:06 PM, Hilmar Lapp hlapp@nescent.org wrote:
Hi Steve and Cam,
I have a question re: your design of the current Semantic Darwin Core ontology. (And this is assuming that with the published 0.2 version I have the latest one in hand.)
My understanding is that all classes in DSW are imported from either DwC or DC (or FOAF), in order to reuse those terms. While that's a good idea in principle, it seems that DSW is actually not saying much new (in a semantic sense) about them, except to declare them to be OWL classes, and to assert them as disjoint from each other (or equivalent in one case). DSW then adds a variety of object (and some data) properties, which distinguish themselves from those in DwC by declaring domain and range axioms for them. But that doesn't say anything about the classes either, nor does it, I would argue, about the properties - domain and range constraint really only say something about the instances for which one asserts those properties.
So by itself the DSW won't allow me to infer anything about the classes and properties in the ontology (aside from disjointness), though it will allow me to make more inferences about instance data to which it is applied than DwC would. And those additional inferences would consist only of the instances' class memberships (and their non-memberships).
I'm wondering a) whether I'm missing something here and am in error, and if not, b) whether the above was indeed the extent of what you wanted to achieve with DSW. Either way, what are your current plans with the ontology? It doesn't seem to have changed for a while.
(And please forgive me if this isn't the right list to post to - I couldn't find a DSW-specific one on the Google code homepage.)
Cheers,
-hilmar
=========================================================== : Hilmar Lapp -:- Durham, NC -:- informatics.nescent.org : ===========================================================
tdwg-content mailing list tdwg-content@lists.tdwg.org http://lists.tdwg.org/mailman/listinfo/tdwg-content
On Jan 25, 2012, at 5:30 PM, Bob Morris wrote:
my own present position is that reasoning on data described by an ontology is way more interesting than reasoning on the ontology itself
I actually don't disagree with that. But the degree to which it is interesting still depends on how richly you can reason. I'm not saying that the ability to infer that an element of the data is an instance of one class and not an instance of 10 other classes isn't useful in some cases, but by itself that's also not very rich yet.
If, however, our ontology contained assertions about the members of a class, for example in the form of property restrictions, or if classes are in rich subclass and/or partonomy hierarchies, you could infer more about those instances.
Just for clarification for the Bob Morrises among us who enjoy jumping a step ahead :-) I myself am right now just trying to ascertain whether my understanding of DSW is correct or not.
-hilmar
Dear Hilmar,
Thanks for your questions about DSW. Steve will have more to add, but the simple answer is that DSW was not indented to say anything new about the existing DwC classes themselves, other than offering a suggestion, based on Steve's extensive search for community consensus on usage in the tdwg-content list, of how the classes best relate to one another. These relations are indicated by the coining of a set of predicates that offer more semantic content than the generic dwc:relatedResourceID, and permit more succinct SPARQL searches, as Bob pointed out.
Based on Steve's review, the range of ways of using the dwc:Occurrence class has been wide, and we suggested a restricted usage in this ontology: the documented presence of an individual organism at a particular event (= space x time); a specimen/photo/observation is in this case not the occurrence itself, but provides evidence for the occurrence.
This all depends on the one new class in DSW, the IndividualOrganism, which Steve and others have been proposing as a fundamental class for modeling biodiversity data. With an IndividualOrganism class, we can easily link from the knowledge domain of biological specimens to that of population biology, where observed/remeasured individuals are the core unit.
We developed DSW to serve our pragmatic need for a semantic template with which to serve data as RDF. Reasoning with it is possible, as you and Bob noted, but I agree, the range of discoveries is limited, because of the few logical restrictions currently in DSW. Perhaps we should not have used the word `ontology' to describe it?
As is still the case now as when Steve announced DSW to tdwg-content, we consider DSW primarily a suggestion for further discussion, and hopefully for further community development (i.e. via the nascent TDWG RDF/OWL Task Group). It `makes sense' to us, and we're using it to model data, but would appreciate significant comment and criticism, including the need to add more logical restrictions.
Best,
Cam
[ There's more on our rationale behind DSW at: http://code.google.com/p/darwin-sw/wiki/Rationale ]
On Wed, 25 Jan 2012, Hilmar Lapp wrote:
Hi Steve and Cam,
I have a question re: your design of the current Semantic Darwin Core ontology. (And this is assuming that with the published 0.2 version I have the latest one in hand.)
My understanding is that all classes in DSW are imported from either DwC or DC (or FOAF), in order to reuse those terms. While that's a good idea in principle, it seems that DSW is actually not saying much new (in a semantic sense) about them, except to declare them to be OWL classes, and to assert them as disjoint from each other (or equivalent in one case). DSW then adds a variety of object (and some data) properties, which distinguish themselves from those in DwC by declaring domain and range axioms for them. But that doesn't say anything about the classes either, nor does it, I would argue, about the properties - domain and range constraint really only say something about the instances for which one asserts those properties.
So by itself the DSW won't allow me to infer anything about the classes and properties in the ontology (aside from disjointness), though it will allow me to make more inferences about instance data to which it is applied than DwC would. And those additional inferences would consist only of the instances' class memberships (and their non-memberships).
I'm wondering a) whether I'm missing something here and am in error, and if not, b) whether the above was indeed the extent of what you wanted to achieve with DSW. Either way, what are your current plans with the ontology? It doesn't seem to have changed for a while.
(And please forgive me if this isn't the right list to post to - I couldn't find a DSW-specific one on the Google code homepage.)
Cheers,
-hilmar
=========================================================== : Hilmar Lapp -:- Durham, NC -:- informatics.nescent.org : ===========================================================
Hilmar, Cam summarized the situation with DSW (http://code.google.com/p/darwin-sw/) pretty accurately. Additional comments:
It has been suggested that the way forward on Darwin Core is not to try to rewrite the standard to do everything that everyone wants, but to build on it in a layered fashion. The first layer might be to simply define terms for the purpose of standardization. The second layer might be to nail down the relationships among the class and property terms of the first layer and to establish how they could be represented in RDF to allow for data transmission. A third layer (based on the framework laid out in the second layer) might be to describe more complex semantics that could be used for reasoning. As I've described this, the first layer is already in place in the form of the current standard. DSW is an attempt to create a second layer. Although we didn't put much in the way of complex semantics into DSW, we hoped that we did not do anything that would prevent someone from using it as the basis for a third layer.
Again, recapitulating what others have said previously, there wouldn't necessarily have to be single "upper" layers on this cake that I described. Particularly on layer three, people might create different ontologies to accomplish different things. But minimizing the number of second layers to the smallest possible number would be desirable if the goal is data interchange. That's one reason why I took great pains to try to document what I thought I was hearing as far as community consensus was concerned. You can look at the various wiki pages to judge for your self the extent to which I was successful. There were certain decisions that Cam and I made (such as creating those object properties to connect the DwC classes rather than trying to make use of the various existing DwC ID terms, and assigning ranges and domains to some of the terms) which may or may not have been good ideas. As Cam said, we were hoping for discussion and constructive criticism on our approach, which we would still welcome.
Bob was correct in his post that I (but perhaps not Cam) have pretty much reached my limits as far as my ability to add meaningfully to what we've put into DSW (and Bob was also basically correct in his outline of design criteria). DSW does for me what I need - provides a way to structure RDF describing my images. At this point I'm trying to focus my efforts on the work of the TDWG RDF/OWL task group, in particular coming up with a beginner's guide to RDF and to try to compile a list of issues that have been raised in the past to complement whatever Joel has gotten from his survey. So I probably won't be putting much further effort into DSW development.
Steve
Cam Webb wrote:
Dear Hilmar,
Thanks for your questions about DSW. Steve will have more to add, but the simple answer is that DSW was not indented to say anything new about the existing DwC classes themselves, other than offering a suggestion, based on Steve's extensive search for community consensus on usage in the tdwg-content list, of how the classes best relate to one another. These relations are indicated by the coining of a set of predicates that offer more semantic content than the generic dwc:relatedResourceID, and permit more succinct SPARQL searches, as Bob pointed out.
Based on Steve's review, the range of ways of using the dwc:Occurrence class has been wide, and we suggested a restricted usage in this ontology: the documented presence of an individual organism at a particular event (= space x time); a specimen/photo/observation is in this case not the occurrence itself, but provides evidence for the occurrence.
This all depends on the one new class in DSW, the IndividualOrganism, which Steve and others have been proposing as a fundamental class for modeling biodiversity data. With an IndividualOrganism class, we can easily link from the knowledge domain of biological specimens to that of population biology, where observed/remeasured individuals are the core unit.
We developed DSW to serve our pragmatic need for a semantic template with which to serve data as RDF. Reasoning with it is possible, as you and Bob noted, but I agree, the range of discoveries is limited, because of the few logical restrictions currently in DSW. Perhaps we should not have used the word `ontology' to describe it?
As is still the case now as when Steve announced DSW to tdwg-content, we consider DSW primarily a suggestion for further discussion, and hopefully for further community development (i.e. via the nascent TDWG RDF/OWL Task Group). It `makes sense' to us, and we're using it to model data, but would appreciate significant comment and criticism, including the need to add more logical restrictions.
Best,
Cam
[ There's more on our rationale behind DSW at: http://code.google.com/p/darwin-sw/wiki/Rationale ]
On Wed, 25 Jan 2012, Hilmar Lapp wrote:
Hi Steve and Cam,
I have a question re: your design of the current Semantic Darwin Core ontology. (And this is assuming that with the published 0.2 version I have the latest one in hand.)
My understanding is that all classes in DSW are imported from either DwC or DC (or FOAF), in order to reuse those terms. While that's a good idea in principle, it seems that DSW is actually not saying much new (in a semantic sense) about them, except to declare them to be OWL classes, and to assert them as disjoint from each other (or equivalent in one case). DSW then adds a variety of object (and some data) properties, which distinguish themselves from those in DwC by declaring domain and range axioms for them. But that doesn't say anything about the classes either, nor does it, I would argue, about the properties - domain and range constraint really only say something about the instances for which one asserts those properties.
So by itself the DSW won't allow me to infer anything about the classes and properties in the ontology (aside from disjointness), though it will allow me to make more inferences about instance data to which it is applied than DwC would. And those additional inferences would consist only of the instances' class memberships (and their non-memberships).
I'm wondering a) whether I'm missing something here and am in error, and if not, b) whether the above was indeed the extent of what you wanted to achieve with DSW. Either way, what are your current plans with the ontology? It doesn't seem to have changed for a while.
(And please forgive me if this isn't the right list to post to - I couldn't find a DSW-specific one on the Google code homepage.)
Cheers,
-hilmar
=========================================================== : Hilmar Lapp -:- Durham, NC -:- informatics.nescent.org : ===========================================================
.
Steve & Cam,
Thanks for your responses, very helpful. It does seem that at least I wasn't making gross errors in my understanding of the current DSW.
Just to give some context, I'm trying to prototype some ways of using semantic web standards to formally assert relationships between concepts in biodiversity specimen record annotation, and those in annotating metagenomics sequence experiments (or results). I want to use DSW on one side for this, in order to have a stringent representation of the key concepts, and I am creating an OWL ontology corresponding to GCDML for the other side.
-hilmar
Sent with a tap.
On Jan 26, 2012, at 7:49 PM, Steve Baskauf steve.baskauf@vanderbilt.edu wrote:
Hilmar, Cam summarized the situation with DSW (http://code.google.com/p/darwin-sw/) pretty accurately. Additional comments:
It has been suggested that the way forward on Darwin Core is not to try to rewrite the standard to do everything that everyone wants, but to build on it in a layered fashion. The first layer might be to simply define terms for the purpose of standardization. The second layer might be to nail down the relationships among the class and property terms of the first layer and to establish how they could be represented in RDF to allow for data transmission. A third layer (based on the framework laid out in the second layer) might be to describe more complex semantics that could be used for reasoning. As I've described this, the first layer is already in place in the form of the current standard. DSW is an attempt to create a second layer. Although we didn't put much in the way of complex semantics into DSW, we hoped that we did not do anything that would prevent someone from using it as the basis for a third layer.
Again, recapitulating what others have said previously, there wouldn't necessarily have to be single "upper" layers on this cake that I described. Particularly on layer three, people might create different ontologies to accomplish different things. But minimizing the number of second layers to the smallest possible number would be desirable if the goal is data interchange. That's one reason why I took great pains to try to document what I thought I was hearing as far as community consensus was concerned. You can look at the various wiki pages to judge for your self the extent to which I was successful. There were certain decisions that Cam and I made (such as creating those object properties to connect the DwC classes rather than trying to make use of the various existing DwC ID terms, and assigning ranges and domains to some of the terms) which may or may not have been good ideas. As Cam said, we were hoping for discussion and constructive criticism on our approach, which we would still welcome.
Bob was correct in his post that I (but perhaps not Cam) have pretty much reached my limits as far as my ability to add meaningfully to what we've put into DSW (and Bob was also basically correct in his outline of design criteria). DSW does for me what I need - provides a way to structure RDF describing my images. At this point I'm trying to focus my efforts on the work of the TDWG RDF/OWL task group, in particular coming up with a beginner's guide to RDF and to try to compile a list of issues that have been raised in the past to complement whatever Joel has gotten from his survey. So I probably won't be putting much further effort into DSW development.
Steve
Cam Webb wrote:
Dear Hilmar,
Thanks for your questions about DSW. Steve will have more to add, but the simple answer is that DSW was not indented to say anything new about the existing DwC classes themselves, other than offering a suggestion, based on Steve's extensive search for community consensus on usage in the tdwg-content list, of how the classes best relate to one another. These relations are indicated by the coining of a set of predicates that offer more semantic content than the generic dwc:relatedResourceID, and permit more succinct SPARQL searches, as Bob pointed out.
Based on Steve's review, the range of ways of using the dwc:Occurrence class has been wide, and we suggested a restricted usage in this ontology: the documented presence of an individual organism at a particular event (= space x time); a specimen/photo/observation is in this case not the occurrence itself, but provides evidence for the occurrence.
This all depends on the one new class in DSW, the IndividualOrganism, which Steve and others have been proposing as a fundamental class for modeling biodiversity data. With an IndividualOrganism class, we can easily link from the knowledge domain of biological specimens to that of population biology, where observed/remeasured individuals are the core unit.
We developed DSW to serve our pragmatic need for a semantic template with which to serve data as RDF. Reasoning with it is possible, as you and Bob noted, but I agree, the range of discoveries is limited, because of the few logical restrictions currently in DSW. Perhaps we should not have used the word `ontology' to describe it?
As is still the case now as when Steve announced DSW to tdwg-content, we consider DSW primarily a suggestion for further discussion, and hopefully for further community development (i.e. via the nascent TDWG RDF/OWL Task Group). It `makes sense' to us, and we're using it to model data, but would appreciate significant comment and criticism, including the need to add more logical restrictions.
Best,
Cam
[ There's more on our rationale behind DSW at: http://code.google.com/p/darwin-sw/wiki/Rationale ]
On Wed, 25 Jan 2012, Hilmar Lapp wrote:
Hi Steve and Cam,
I have a question re: your design of the current Semantic Darwin Core ontology. (And this is assuming that with the published 0.2 version I have the latest one in hand.)
My understanding is that all classes in DSW are imported from either DwC or DC (or FOAF), in order to reuse those terms. While that's a good idea in principle, it seems that DSW is actually not saying much new (in a semantic sense) about them, except to declare them to be OWL classes, and to assert them as disjoint from each other (or equivalent in one case). DSW then adds a variety of object (and some data) properties, which distinguish themselves from those in DwC by declaring domain and range axioms for them. But that doesn't say anything about the classes either, nor does it, I would argue, about the properties - domain and range constraint really only say something about the instances for which one asserts those properties.
So by itself the DSW won't allow me to infer anything about the classes and properties in the ontology (aside from disjointness), though it will allow me to make more inferences about instance data to which it is applied than DwC would. And those additional inferences would consist only of the instances' class memberships (and their non-memberships).
I'm wondering a) whether I'm missing something here and am in error, and if not, b) whether the above was indeed the extent of what you wanted to achieve with DSW. Either way, what are your current plans with the ontology? It doesn't seem to have changed for a while.
(And please forgive me if this isn't the right list to post to - I couldn't find a DSW-specific one on the Google code homepage.)
Cheers,
-hilmar
=========================================================== : Hilmar Lapp -:- Durham, NC -:- informatics.nescent.org : ===========================================================
.
-- Steven J. Baskauf, Ph.D., Senior Lecturer Vanderbilt University Dept. of Biological Sciences
postal mail address: VU Station B 351634 Nashville, TN 37235-1634, U.S.A.
delivery address: 2125 Stevenson Center 1161 21st Ave., S. Nashville, TN 37235
office: 2128 Stevenson Center phone: (615) 343-4582, fax: (615) 343-6707 http://bioimages.vanderbilt.edu
participants (4)
-
Bob Morris -
Cam Webb -
Hilmar Lapp -
Steve Baskauf