uima-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Peter Kl├╝gl <peter.klu...@averbis.com>
Subject Re: Ruta COUNT issue
Date Wed, 16 Aug 2017 06:37:36 GMT
Hi,


I haven't tested the rule for bugs yet , so here's just a comment for now.

The order of evaluation of the conditions is fixed. The usage of 
conditions on a wildcard can lead to unexpected matches (and it can 
considerbaly decrease the runtime performance).

There is a difference between a wildcard and a reluctant quantfier like 
"ANY+?".

The wildcard takes a look at the next rule element and then evaluates 
its conditons. If a condition fails, the next rule element is consulted 
again, like moving forward on its iterator, and then the conditions are 
checked again. With a normal matching rule, the conditions are directly 
evaluated and the next element is not considered anew or skipped.


I have to test the rule in order to give you a specific answer, but that 
will take some time (I am a bit busy right now)


Best,


Peter


Am 11.08.2017 um 22:23 schrieb Tom Vacek:
> Assume I have made PositiveContext and NegativeContext annotations from a
> word list.   I would like a wildcard rule that allows any number of
> PositiveContext tokens, some (eg 4) unknown tokens, but no NegativeContext
> tokens.
> I have this rule:
> INT contextCount;
> (Keyword1  #{COUNT(PositiveContext,0,100,contextCount),
> CONTAINS(W,0,contextCount+4), -CONTAINS(NegativeContext)} Keyword2
> The rule behaves unpredictably, perhaps as if the order of evaluation of
> the conditions is not fixed.  Is this the case?  Or is it some other bug?
>
> Thanks,
>
> Tom
>


Mime
View raw message