Overloading OWL sameAs starts with:
Description: General Issue: owl:sameAs is being used in the linked data community in a way that is inconsistent with its semantics.
Read the document but in summary: People use OWL sameAs to mean different things.
I don’t see how their usage can be “inconsistent with its semantics.”
Words don’t possess self-executing semantics that bind us. Rather the other way round I think.
If OWL sameAs had some “original” semantic, it changed by the process of semantic drift.
Semantic drift is where the semantics of a token changes over time or across communities due to its use by people.
URIs or tokens may be “stable,” but the evidence is that the semantics of URIs or tokens are not.
The question is how to manage changing, emerging, drifting semantics? (Not a question answered by a static semantic model of URI based identity.)
PS: RDF researchers have recognized semantic drift and have proposed solutions for addressing it. More on that anon.
Questions:
- Select a classification more than 30 years old and randomly select one book for each 5 year period for the last 30 years. What (if any) semantic drift do you see in the use of this classification?
- Exchange your list with a classmate. Do you agree/disagree with their evaluation? Why?
- Repeat the exercise in #1 and #2 but use a classification where you can find books between 30 and 60 years ago. Select one book per 5 year period.
Entropy increases, and so does semantic entropy.
If the semantic of a URI (or any other sign) were a drifting boat, we might be able to come up with a mechanism that could track its drift. But the situation is much more desperate: all allegedly seaworthy vessels inexorably dissolve into chunks that themselves drift and/or dissolve. It’s normal for “sameAs”, like every other sign, to become the moniker for the scattered remains of what was once (at least allegedly) a single semantic.
We must use signs, or we can’t communicate. But if our strategies for supporting and prolonging the existence of civilization are based on false estimates of the permanence of the mappings between signs and their semantics, then civilization is surely doomed.
Comment by Steve Newcomb — October 8, 2010 @ 12:51 pm
About a decade ago when ontologies were discovered widely in computer science there was a serious concern about overloading is-a relation. People used is-a relation inconsistently in their ontologies. To me, there is a striking analogy between current debate about inconsistent use of sameAs and past debate about is-a. If I recall it right, past debate about is-a raised people’s awareness of the problem and people started really thinking about their relations.
Patrick, I totally agree with you about the semantic drift. Semantic is “injected” to the words when we use them and semantic changes over time. Therefore, I think, we should use words and discuss about their meaning in order to preserve mutual understanding. If there is a way to model semantic drift, I sure would like to know more. Somehow I have a feeling it is realized by introducing new relations that tie different flavors of the original relation :).
Comment by Aki — October 8, 2010 @ 3:47 pm
I do get the impression that human inconsistency in usage plays a large part in the semantic drift of owl:sameAs. As people create data, they may apply ontologies inconsistently (exactly because words don’t have fixed meanings their usage by different people triggers inconsistency). It also seems impossible to foresee all potential confusions and disambiguate in advance so as to retain “stable” semantics over time.
Comment by Saskia — October 11, 2010 @ 4:58 am
Saskia – As an RDF advocate told me when I pointed out that people weren’t going to be consistent, “that’s a user problem.”
But we are developing systems to be used by….people. We can develop ideal systems, completely logically consistent, but if they aren’t usable by people or do what people what, what good are they?
Impossible for foresee all changes in semantics but more coming this week on possible approaches to managing changes in semantics.
Aki – Curious, what do you think of the usage of is-a now? Is it overloaded in ontology usage? What about in topic maps?
Comment by Patrick Durusau — October 11, 2010 @ 5:14 am
Patrick, I am not a professional ontology builder nor have watched the field lately but I have an impression that people rarely use plain is-a relation in their ontologies today. I have a feeling that people use today relations such as superclass-subclass, class-instance, type-of, related-to instead of the is-a relation. In my opinion, this may be the destiny of Owl’s sameAs relation too. The relation will be split into several other relations that capture the observed flavors of current usage of sameAs. If you are interested in examining the debate about is-a overloading, I recall that at least Nicola Guarino did write an interesting article about it… seems to be online at
http://webcache.googleusercontent.com/search?q=cache:Qhsy05FxpVsJ:citeseerx.ist.psu.edu/viewdoc/download%3Fdoi%3D10.1.1.18.1288%26rep%3Drep1%26type%3Dps+overloading+is-a+relation+Guarino&cd=1&hl=en&ct=clnk
Comment by Aki — October 11, 2010 @ 7:03 am
[…] the semantic vagaries of owl:sameAs (Semantic Drift and Linked Data/Semantic Web), I have to wonder about the longer term maintenance of owl:sameAs […]
Pingback by Hafslund Sesam — an archive on semantics « Another Word For It — June 13, 2013 @ 12:27 pm