How large is the OWL ontology that an OWL reasoner can deal with?

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

How large is the OWL ontology that an OWL reasoner can deal with?

Ling
Hi, all:

Does anyone of you have experience on reasoning with large OWL ontologies? Let's say if I have tens of millions of objects and relations in the ontology, can the reasoner (i.e.Hermit) comfortably reason with?

From a practical point of view, this is an import question.

Thanks.

Ling


_______________________________________________
protege-user mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/protege-user
Reply | Threaded
Open this post in threaded view
|

Re: How large is the OWL ontology that an OWL reasoner can deal with?

Phillip Lord
It's easy to test. The answer is, it depends.

If the ontology is structurally simple (say, classes and subclasses) then a reasoner can cope, although for tens of millions you may need a fair bit of memory. If the ontology is structurally more complex, then the reasoner will fail earlier. And, if you really try hard, you might be able to get them to fail on even a small ontology.

In general, though, I would say that the most ontologies come in quite a bit small than tens of millions of objects.



________________________________________
From: protege-user <[hidden email]> on behalf of Ling <[hidden email]>
Sent: 23 December 2017 20:12
To: User support for WebProtege and Protege Desktop
Subject: [protege-user] How large is the OWL ontology that an OWL reasoner      can deal with?

Hi, all:

Does anyone of you have experience on reasoning with large OWL ontologies? Let's say if I have tens of millions of objects and relations in the ontology, can the reasoner (i.e.Hermit) comfortably reason with?

From a practical point of view, this is an import question.

Thanks.

Ling

_______________________________________________
protege-user mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/protege-user
Reply | Threaded
Open this post in threaded view
|

Re: How large is the OWL ontology that an OWL reasoner can deal with?

Ling
Hi, Phillip:

Typically, what types of axioms could make the ontology very complex and unable to reason? 

Thanks,
Ling

On Sat, Dec 23, 2017 at 2:45 PM, Phillip Lord <[hidden email]> wrote:
It's easy to test. The answer is, it depends.

If the ontology is structurally simple (say, classes and subclasses) then a reasoner can cope, although for tens of millions you may need a fair bit of memory. If the ontology is structurally more complex, then the reasoner will fail earlier. And, if you really try hard, you might be able to get them to fail on even a small ontology.

In general, though, I would say that the most ontologies come in quite a bit small than tens of millions of objects.



________________________________________
From: protege-user <[hidden email]> on behalf of Ling <[hidden email]>
Sent: 23 December 2017 20:12
To: User support for WebProtege and Protege Desktop
Subject: [protege-user] How large is the OWL ontology that an OWL reasoner      can deal with?

Hi, all:

Does anyone of you have experience on reasoning with large OWL ontologies? Let's say if I have tens of millions of objects and relations in the ontology, can the reasoner (i.e.Hermit) comfortably reason with?

From a practical point of view, this is an import question.

Thanks.

Ling

_______________________________________________
protege-user mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/protege-user


_______________________________________________
protege-user mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/protege-user
Reply | Threaded
Open this post in threaded view
|

Re: How large is the OWL ontology that an OWL reasoner can deal with?

Lorenz Buehmann

On 24.12.2017 06:08, Ling wrote:
> Typically, what types of axioms could make the ontology very complex
> and unable to reason? 
That's also something that is not fully predictable. And in general,
it's usually a combination of axiom types. The first entry point would
be to understand the tableau algorithm which is usually used for OWL 2
DL reasoning. Here, the most crucial part is non-determinism, introduced
by several OWL constructs like unions or cardinality restrictions.

An indicator **could** be the theoretical complexity of the underlying
Description Logic [3] , but that's indeed just theory. Note, from this
point of view, OWL 2 DL reasoning (SROIQ would be the Description Logic)
has 2-NExpTime complexity, thus, it should more or less never work. The
good thing is, there are a lots of optimizations which make OWL DL
reasoning usable in practice.

To sum up, this topic is too complex and is ongoing research [1] [2],
but in the end, nobody would ever be able to give you proper metrics
which can tell you whether it works or not. As a hint, "tens of
millions" is quite large for any tableau-based reasoner nowadays. That's
for sure more a task for scalable rule-engines with more light-weight
reasoning support (e.g. OWL horst or just RDFS).


Cheers,
Lorenz


[1] https://dl.acm.org/citation.cfm?id=3032020
[2] https://link.springer.com/content/pdf/10.1007/978-3-642-35176-1_13.pdf
[3] http://www.cs.man.ac.uk/~ezolin/dl/
_______________________________________________
protege-user mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/protege-user
Reply | Threaded
Open this post in threaded view
|

Re: How large is the OWL ontology that an OWL reasoner can deal with?

Ling
Hi, Lorenz:

Thanks for the very good info. and references. In Protege when I create an ontology, how do I know which version of the OWL ontology I am creating, OWL DL, QL, EL and RL? 

And secondly, for a typical inference problem, like:

bornIn(Obama, Hawalli) && partOf(Hawaii, USA)
=>
bornIn(Obama, USA)

which OWL ontology is needed for this inference? 

Happy Holiday.
Ling

On Sat, Dec 23, 2017 at 10:04 PM, Lorenz Buehmann <[hidden email]> wrote:

On 24.12.2017 06:08, Ling wrote:
> Typically, what types of axioms could make the ontology very complex
> and unable to reason? 
That's also something that is not fully predictable. And in general,
it's usually a combination of axiom types. The first entry point would
be to understand the tableau algorithm which is usually used for OWL 2
DL reasoning. Here, the most crucial part is non-determinism, introduced
by several OWL constructs like unions or cardinality restrictions.

An indicator **could** be the theoretical complexity of the underlying
Description Logic [3] , but that's indeed just theory. Note, from this
point of view, OWL 2 DL reasoning (SROIQ would be the Description Logic)
has 2-NExpTime complexity, thus, it should more or less never work. The
good thing is, there are a lots of optimizations which make OWL DL
reasoning usable in practice.

To sum up, this topic is too complex and is ongoing research [1] [2],
but in the end, nobody would ever be able to give you proper metrics
which can tell you whether it works or not. As a hint, "tens of
millions" is quite large for any tableau-based reasoner nowadays. That's
for sure more a task for scalable rule-engines with more light-weight
reasoning support (e.g. OWL horst or just RDFS).


Cheers,
Lorenz


[1] https://dl.acm.org/citation.cfm?id=3032020
[2] https://link.springer.com/content/pdf/10.1007/978-3-642-35176-1_13.pdf
[3] http://www.cs.man.ac.uk/~ezolin/dl/
_______________________________________________
protege-user mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/protege-user


_______________________________________________
protege-user mailing list
[hidden email]
https://mailman.stanford.edu/mailman/listinfo/protege-user