Skip to content

Conversation

@danyaljj
Copy link
Member

Create a pairwise classifier and add a constrained classifier on top of it.
** Not carefully tested. Under construction **

FYI @bhargav

@danyaljj
Copy link
Member Author

Did a bit of work and fixed some issues. The inference has some problems and I need to spend more time debugging them.

@kordjamshidi
Copy link
Member

kordjamshidi commented Apr 18, 2016

where is the code? I can look at it. Specially the way you write constraints is tricky.

@danyaljj
Copy link
Member Author

On my fork; branch "danyaljj:pairwise-pos2".
Run the POS app.

}

def sentenceLabelsMatch = ConstrainedClassifier.constraint[Sentence] { s: Sentence =>
val constituents = CommonSensors.getPOSConstituents(s.getSentenceConstituent.getTextAnnotation)
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

here

khashab2 added 4 commits April 25, 2016 16:39
# Conflicts:
#	saul-core/src/main/scala/edu/illinois/cs/cogcomp/saul/classifier/ConstrainedClassifier.scala
#	saul-examples/src/main/scala/edu/illinois/cs/cogcomp/saulexamples/nlp/POSTagger/POSClassifiers.scala
#	saul-examples/src/main/scala/edu/illinois/cs/cogcomp/saulexamples/nlp/POSTagger/POSTaggerApps.scala
// posLabelPairs._exists { case (l1, l2) => posClassifierLabelCompatible(c1, c2, l1, l2) }
// }
//constituents.toList._forall { c => posLabels._exists { l => (POSMixedClassifier on c).is(l) } }
new FirstOrderConstant(true)
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@bhargav so I fixed the bug you pointer out but didn't help .. :-/
Another thing is I tested by replacing it with a trivial constraints new FirstOrderConstant(true) which should always be feasible; still returning a weird result ...

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants