Your rule doesn't look right to me, even after the fix, provided I've understood
everything correctly. (I'm assuming that a NamedEntity is a range in a
text and ranges don't overlap.) I would have thought that the distance
between two nonoverlapping ranges is given by
function int d( NamedEntity x, NamedEntity y ){
return Math.min( Math.abs( x.getBeg() - y.getEnd() ),
Math.abs( y.getBeg() - x.getEnd() ) );
}
and that such d-values need to be compared to guarantee the minimum,
taking care to replace NamedEntities with equal contents( "anatomy",
etc) when comparing one distance to another.
-W
On 26/02/2013, Bojan Janisch <bojan.janisch(a)scai.fraunhofer.de> wrote:
Hey again,
I've solved the problem. Seems that the error came from an unhandled
condition.
I hope that the following example helps other people with similar problems:
//Checking if there is and Anatomy-Attribute
$ne1:NamedEntity(
ct1:coveredText,
q1:QuestionId,
q1.contains("anatomy"),
label1:ResponseAttribute,
ne1_begin:begin,
ne1_end:end)
//Checking if there is also a BodySide-Attribute
$ne2:NamedEntity(
ct2:coveredText,
q2:QuestionId,
q2.contains("bodySide"),
label2:ResponseAttribute,
ne2_begin:begin,
ne2_end:end)
//Checking that there is no other Attribute nearer
not $ne3:NamedEntity(
q3:QuestionId,
q3.contains("bodySide") || q3.contains("anatomy"),
ne3_begin:begin,
ne3_end:end,
(ne3_begin > ne1_end && (ne3_begin - ne1_end) < (ne1_begin - ne2_end)) ||
(ne3_end < ne1_begin && (ne1_begin - ne3_end) < (ne2_begin - ne1_end)))
The last condition of ne3 does not handel the fact that ne1 is before ne2.
And I'm also not testing for this fact otherwise. So I changed the last
condition into:
not $ne3:NamedEntity(
q3:QuestionId,
q4.contains(".bodySide") || q4.contains("anatomy"),
ne3_begin:begin,
ne3_end:end,
ne3_begin > ne1_end && ((ne1_end - ne3_begin) < (ne1_end - ne2_begin)) ||
ne3_end < ne1_begin && ((ne1_begin - ne3_end) < (ne1_begin - ne2_end)))
Now it is not allowed that a named entity with questionId of anatomy or
bodyside is
between ne1 and ne2 (doesn't matter which comes first). The runtime
exception did not
occur anymore after I changed the rule.
So seems that the null accessor was a result of an unhandled condition.
Greetings
Bojan
----- Ursprüngliche Mail -----
Von: "Bojan Janisch" <bojan.janisch(a)scai.fraunhofer.de>
An: "Rules Users List" <rules-users(a)lists.jboss.org>
Gesendet: Montag, 25. Februar 2013 18:11:46
Betreff: Runtime Exception null accessor
Hello Guys,
I've thought many hours about this problem, but don't know how to fix it.
I've also searched the web, getting only some unfinished problems. Maybe
someone here knows how to solve it:
Each time I'm running JDrools through my textfiles (I'm using it in
textmining concepts), I get the same runtime exception:
25.02.2013 18:03:59
org.apache.uima.collection.impl.cpm.engine.CPMThreadGroup process
SCHWERWIEGEND: The CPM thread group caught the following unhandled error:
java.lang.RuntimeException (Thread Name: Thread-5)
25.02.2013 18:03:59
de.fraunhofer.scai.bio.uima.core.deploy.StatusCallbackListenerImpl
entityProcessComplete(151)
SCHWERWIEGEND: Exception occurred
java.lang.RuntimeException: Null accessor on node: a1_begin
at
de.fraunhofer.scai.bio.uima.core.deploy.StatusCallbackListenerImpl.entityProcessComplete(StatusCallbackListenerImpl.java:151)
at
org.apache.uima.collection.impl.cpm.engine.CPMThreadGroup.notifyListener(CPMThreadGroup.java:103)
at
org.apache.uima.collection.impl.cpm.engine.CPMThreadGroup.uncaughtException(CPMThreadGroup.java:86)
at java.lang.Thread.dispatchUncaughtException(Thread.java:1888)
For me it looks like an unhandeld race condition (but I'm not quite sure).
Does someone have had this or a similar problem regarding null accessors at
runtime?
Thanks for your help.
Bojan
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users