Posts:
27
Registered:
Jun 15, 2009 07:52:26 AM
|
|
|
|
lemmaEntries feature in ClauseEndingPunctuation Type
Posted:
Oct 28, 2009 05:47:41 AM
|
|
|
|
Hi,
I've tested my pear in uima sdk 2.2.2
There were not any errors in LWR while testing documents.
Most of annotations work in the pear installed into uima, but what does that message mean:
28.10.2009 11:51:21 com.ibm.talent.fstSymbols.MiniTestLemmaString test(161)
WARNING: Lemma test failed as Annotation of FS Type com.ibm.langware.uimatypes.ClauseEndingPunctuation expected to have lemmaEntries feature for match.
That prevents to get the most interesting annotations, can I have an influence on this type?
Regards,
Valentin
|
|
Posts:
72
Registered:
Feb 03, 2009 10:46:24 AM
|
|
|
|
Re: lemmaEntries feature in ClauseEndingPunctuation Type
Posted:
Oct 29, 2009 10:55:21 AM
in response to: VVTs's post
|
|
|
|
Hi Valentin. The cause of this issue is that you are testing the "lemma" value on Tokens, but some Tokens (any ClauseEndingPunctuation, for example) do not have a lemma assigned. All the warning is saying is that since no lemma is assigned, the lemma test cannot ever match.
This is nothing to be worried about - we have actually removed this warning from the the product in our internal builds, as we spotted this issue before now, and realised that it is confusing.
I hope this answers your question.
Kevin
|
|
Posts:
27
Registered:
Jun 15, 2009 07:52:26 AM
|
|
|
|
Re: lemmaEntries feature in ClauseEndingPunctuation Type
Posted:
Oct 30, 2009 04:00:24 AM
in response to: VVTs's post
|
|
|
|
Hi, Kevin!
Thank you for the fast and clear reply.
You are right, but let's see futher. Since the first post I've found that rule which is guilty of generating of this warning, however it doesn't include punctuation test :( . I don't know how to put picture here, so I'll send it to your e-mail.
Is it possible to make warnings turnable: to suppress those warnings in my own or add to message guilty rule name.
Regards,
Valentin
|
|
Posts:
72
Registered:
Feb 03, 2009 10:46:24 AM
|
|
|
|
Re: lemmaEntries feature in ClauseEndingPunctuation Type
Posted:
Oct 30, 2009 08:23:16 AM
in response to: VVTs's post
|
|
|
|
Hi Valentin. It's a little bit strange that that rule is testing the lemma, since it should only test lemmas for TitlecaseAlphabetic or LowercaseAlphabetic Tokens. If you want you can email me the actual rule grammar (select "Properties Tab -> Show Rule Grammar..." to see this) and I can tell you if there is a problem with how we are writing the rule grammar.
Can I just ask - is there a reason that you need to test the Token values at all? You have Dictionary / Rule annotations covering all annotations. For a rule like this, I'd expect most, if not all of, the "Subtree" nodes to be unchecked. This would ensure that only the most general pattern was tested for. Otherwise, you will only get a very small number of matches - you need to have the exact number of Tokens with the exact lemma you're testing for and this seriously restricts how effective the rule is.
Regards,
Kevin
|
|
Posts:
27
Registered:
Jun 15, 2009 07:52:26 AM
|
|
|
|
Re: lemmaEntries feature in ClauseEndingPunctuation Type
Posted:
Oct 30, 2009 10:38:12 AM
in response to: KevinCunnane's post
|
|
|
|
Kevin,
thanks.
I'll redesign those rules on tuesday.
The reason of checking tokens is that States can consist all CIS countries, and structure of a Minisry name is the same. The Russian Federation documents are the most important among then for us. They are different, so I can make almost the same rule bases for both types. :)
Happy Halloween! :)
Regards,
Valentin
|
|
Posts:
27
Registered:
Jun 15, 2009 07:52:26 AM
|
|
|
|
Re: lemmaEntries feature in ClauseEndingPunctuation Type
Posted:
Nov 03, 2009 09:41:14 AM
in response to: VVTs's post
|
|
|
|
Hi, Kevin!
That's true. Lemma test in the Token underneath the higher-level annotation is the issue in my situation. There were two rules with that test.
Log is clean now.
Thanks,
Valentin
|
|
|
|
