Running PrisM

[sadger]

Hi i'm trying to get a working version of PrisM. I edited the pom.xml as required. However, when I try and compile it with mvn install it generates a jar file successfully. I include the built jar file into my eclipse project as usual but gives an error about missing classes.

Caused by: java.lang.ClassNotFoundException: uk.ac.ox.cs.JRDFox.JRDFStoreException
    at java.net.URLClassLoader.findClass(URLClassLoader.java:381)
    at java.lang.ClassLoader.loadClass(ClassLoader.java:424)
    at sun.misc.Launcher$AppClassLoader.loadClass(Launcher.java:331)
    at java.lang.ClassLoader.loadClass(ClassLoader.java:357)

I suspect the maven built is not producing a self contained jar.
Do you have a precompiled linux jar that I can use or more instructions how to build it correctly?

Thanks.

[anaphylactic]

Hi,
Thanks for downloading PrisM and sorry that you had problems trying to use it! I believe I have now fixed the issue, if you update the project and follow the new instructions in the README file it should work correctly - feel free to let me know if it doesn't!
Cheers,
Ana

[sadger]

Great. Seems to be working now. Thanks for taking the time to make those changes. I'll see if I can get some modules extracted now.

[anaphylactic]

No problem at all! I hope you will find the tool useful ;)

[sadger]

Hi, I have managed to get some module extraction working with PrisM, I am particularly interesting in the model-inseparability relation and seeing how good an approximation PrisM produces. I'm one of the authors from " Lower and Upper Approximations for Depleting Modules of Description Logic Ontologies" in which we compared STAR (\bot\top*) and AMEX (and a combination of both modules) to a lower approximation to estimate how close the modules were to the ideal minimal ones.

One problem I am having with PrisM is extracting modules for several signatures, for a single extraction it works correctly producing a module.

Is it possible to extract multiple modules with a single instance of PrisM without unloading and reloading the ontology? I am getting an OWLOntologyAlreadyExistsException when trying to do the following.

PrisM extractor = new PrisM(ont, InseparabilityRelation.MODEL_INSEPARABILITY);
extractor.extract(signature1,false)
extractor.extract(signature2,false)
extractor.extract(signature3,false)
...

Is there a way to make it work to extract multiple signatures, maybe I am going about it incorrectly?

Thanks again,

William

[anaphylactic]

Hi William,

I am familiar with your work, of course! :)
I think I have fixed that bad behaviour now, if you update the project and rebuild the jar it should work as you need it to. Do let me know if you encounter any other problems.
And thanks for your comments, this helps make the system better!

Cheers,
Ana

[sadger]

Sorry, got the same issue again with the OWLOntologyAlreadyExistsException.
It only happens now if I do extract(signature,false) when it is extract(signature,true) I don't have the issue which is an improvement based on your changes.

[anaphylactic]

That constructor should actually be private and only for testing purposes - I've changed that now.
The first thing that PrisM does is extracting a preliminary syntactic locality based module. That simplifies the work to be done in a cheap way. Calling the extract method with that boolean flag set to false makes this step get skipped, but there is no advantage in that so it doesn't make sense to keep that constructor public. Hope that makes sense!
Ana

[sadger]

Oh ok, that makes sense. I can understand pre-processing with the locality based modules we do it in some of our extraction algorithms too. Is the same module produced with and without the preprocessing step? If it was only intended to be used with the locality preprocessing I will do some investigation using that approach, not a problem. I appreciate you putting the time in to make these changes and answer my queries.

William

[anaphylactic]

It's not a problem at all, it's very nice to have somebody interested in the work! :)
In ontologies with 'right hand side' disjunctions we may get a smaller module with the preprocessing step; in all other ontologies it doesn't make a difference.

[sadger]

Ok great, thanks. I'll give it a go with the preprocessing as you intended then.

[sadger]

Hi, I've unfortunately got a couple more issues that I have discovered while testing PrisM.

1. If the ontology does not have a global ontology IRI or document IRI which is quite common depending how the ontology is generated, PrisM will cause a null pointer exception. I can work around this by assigning the ontology one of course.

2. I am getting some strange output for some ontologies and signatures, axioms that look like
   A ⊑ ∀ r.<internal:def#1>. I have produced a small ALCQ ontology (10 axioms) and signature for which this problem can be reproduced. For comparative purposes the locality STAR module I extracted contains 1 axiom for this signature whereas the datalog generated module (for the model inseparability relation) contains 11 (one more than the entire ontology), including these weird symbols. Does my input ontology have to be in a certain format for PrisM to work?

Ontology: www.csc.liv.ac.uk/~wgatens/small-example.owl
Signature: www.csc.liv.ac.uk/~wgatens/example-sig

[anaphylactic]

I will try to fix the IRI issue ASAP.
The other issue was already in my todo list, but that may take a little longer. This happens because PrisM works with a normalised version of the input ontology and does not currently undo this normalisation. What it outputs is still inseparable from the original ontology w.r.t. the given signature according to the chosen inseparability relation, but it is not, strictly speaking, a module of the original ontology - rather a module of its normalisation.

[sadger]

Ok, I can understand why that would happen if you are normalising first. For comparative purposes I think I will wait until the implementation extracts a module of the original ontology as I cannot accurately compare the sizes of modules, the main metric of interest, at the moment.

Please do not feel the need to fix these things quickly on my behalf. If you just let me know when there is a new version to try I'll happily do some more testing.

Thanks,

William

[anaphylactic]

No worries - all the issues you're mentioning are fair enough and I did need to fix them anyway. And reversing the normalisation really was in my plans too :). I will let you know when I have finished that - thanks for your patience!
Ana

[sadger]

Hi Ana, Hope you are well.

My supervisor brought up an interesting idea, wondering if its feasible. Could one normalise an ontology to work with PrisM then extract a PrisM/locality/AMEX module from the normalisation and for a signature compare the sizes of such modules. The output may still have these fresh symbols but I would hope they are directly comparable, maybe it is not so simple.

Thanks,

William