[Pellet-users] Debugging rules

[Evren Sirin]

On 3/29/08 11:19 PM, Alan Ruttenberg wrote:
> On Mar 28, 2008, at 11:28 PM, Markus Stocker wrote:
>
>   
>> Andrew,
>>
>> As it happens, today I sat in a meeting where part of the discussion
>> was that one of the advantages of DL (tableau) reasoning vs. rule
>> based reasoning is that in DL it is possible to provide explanations
>> while with rules it is hard to explain why. Though I'm going to state
>> this carefully but I believe it's not possible to provide an
>> explanation service for rule based reasoning.
>>     
>
> Not sure how you conclude this. There is a long line of work in this  
> area, e.g. http://www.cogentex.com/papers/explanation-iwnlg98.pdf and  
> AFAIK, much less in the way of generating explanations from DL  
> reasoners (some of the best of which, though,  comes from the Pellet  
> team).
>   

It is certainly true that there is a lot of research done on 
explanations for rule-based systems, especially in the context of expert 
systems. The simplest approach (showing the trace of rule firings) is 
very similar to axiom pinpointing service Pellet provides (showing the 
source axioms causing an inference). However, in practice, expert 
systems tend to mix domain knowledge and control knowledge (meta-rules) 
used to guide the system make inferences efficiently so the explanations 
that come out can be confusing. Rule priorities used in most systems is 
another similar example that makes explanations trickier. This is not to 
say that generating explanations in rule-based systems are impossible 
but to give context for why Marcus was saying it is harder to do so. 
Also, part of the discussion that Marcus mentions was about how DL 
systems can provide explanations for TBox inferences (e.g. concept 
satisfiability) as easily as for ABox inferences whereas such 
explanations are harder for rule-based systems.

> Perhaps you mean that there has not been much work in the area of  
> explaining hybrid DL/Rule systems, which would probably be true.
>   

Yes, this is certainly true. In my reply to Andrew's original message, I 
outlined some possibilities about how this can be done using Pellet's 
explanation feature.

Cheers,
Evren

> -Alan
>
>   
>> Anyway, could you provide a minimal example of your case? I think that
>> would be helpful especially with a simple ontology and a single rule.
>>
>> Cheers,
>> markus
>>
>> On Fri, Mar 28, 2008 at 4:44 PM, Crapo, Andrew (GE, Research)
>> <crapo at ge.com> wrote:
>>     
>>>
>>> I have a simple ontology and a SWRL rule that I am testing. It  
>>> seems to me
>>> that the rule should fire but the conclusion isn't set. How can I  
>>> debug
>>> this? Create some kind of trace of why rules do or do not fire?
>>>
>>> Andrew Crapo
>>> GE Global Research
>>> Information Scientist
>>>
>>> T 518 387 5729
>>> F 518 387 6104
>>> D *833 5729
>>> E crapo at research.ge.com
>>> www.research.ge.com
>>>
>>> One Research Circle
>>> Niskayuna, NY 12309 USA
>>> General Electric Company
>>> _______________________________________________
>>>  Pellet-users mailing list
>>>  Pellet-users at lists.owldl.com
>>>  http://lists.owldl.com/mailman/listinfo/pellet-users
>>>  _______________________________________________
>>>
>>>  Sponsored by Clark & Parsia, LLC http://clarkparsia.com/
>>>
>>>       
>> _______________________________________________
>> Pellet-users mailing list
>> Pellet-users at lists.owldl.com
>> http://lists.owldl.com/mailman/listinfo/pellet-users
>> _______________________________________________
>>
>> Sponsored by Clark & Parsia, LLC http://clarkparsia.com/
>>     
>
> _______________________________________________
> Pellet-users mailing list
> Pellet-users at lists.owldl.com
> http://lists.owldl.com/mailman/listinfo/pellet-users
> _______________________________________________
>
> Sponsored by Clark & Parsia, LLC http://clarkparsia.com/
>