dkpro-wsd icon indicating copy to clipboard operation
dkpro-wsd copied to clipboard

Evaluators fail if the input is not POS-tagged

Open logological opened this issue 10 years ago • 5 comments

Originally reported on Google Code with ID 41

The cluster evaluator assumes the input is POS-tagged, and fails in getTestAndGoldResults()
if it is not.  This is because it stores results in a hash table keyed by part of speech;
since the POS is null we get a NullPointerException.

The problem may exist for other evaluators as well.

Reported by [email protected] on 2013-10-16 10:45:07

logological avatar Jun 24 '15 15:06 logological

Attached test case.

Reported by [email protected] on 2013-10-16 10:49:41


- _Attachment: [ClusterEvaluatorTest.java](https://storage.googleapis.com/google-code-attachments/dkpro-wsd/issue-41/comment-1/ClusterEvaluatorTest.java)_

logological avatar Jun 24 '15 15:06 logological

Problem also affects AbstractSingleExactMatchEvaluator

Reported by [email protected] on 2013-10-17 12:40:35

logological avatar Jun 24 '15 15:06 logological

Possible solution would be to add a new element to the POS enumeration to represent
none/unknown POS.  However, we would need to carefully check the rest of our code to
make sure nothing relies on there being only "real" parts of speech.

Reported by [email protected] on 2013-10-17 12:49:01

  • Labels added: Priority-High
  • Labels removed: Priority-Medium

logological avatar Jun 24 '15 15:06 logological

Reported by [email protected] on 2013-11-20 15:03:32

  • Labels removed: Milestone-Release1.0

logological avatar Jun 24 '15 15:06 logological

Reported by [email protected] on 2014-11-04 11:45:25

  • Labels added: Priority-Critical
  • Labels removed: Priority-High

logological avatar Jun 24 '15 15:06 logological