Thanks Thilo, I appreciate your insight. After testing all of the
example opennlp descriptors, the following situation has become apparent...
Except for OpenNLPSentenceDetector (which does not rely on previous
annotations), it seems that none of the example (opennlp) non-aggregates
are working. (They used to work, but now they don't). However, using
them within aggregate descriptors works fine. For example, the
OpenNLPAggregate descriptor does what it should do, but running the
OpenNLPTokenizer descriptor alone does not annotate at all. I'm guessing
this is a datapath configuration (or resource specification) problem, in
which case I guess I'm on my own. The paths seem to check out though.
Have you encountered this before?
Many thanks,
Sam
Thilo Goetz wrote:
Hm, maybe you're not invoking anything that
actually creates sentences? Simply declaring
sentences as input requirement is not enough:
you need to also invoke the annotator to create
sentences. You need to build an aggregate, see
http://incubator.apache.org/uima/downloads/releaseDocs/2.2.1-incubating/docs/html/tutorials_and_users_guides/tutorials_and_users_guides.html#ugr.tug.aae.building_aggregates
If that's not the problem, we'll need more info
(e.g., the full descriptor you created).
HTH,
Thilo
Sam Fisher wrote:
Hi UIMA Users,
I'm very new to UIMA (started exploring it last week). From
following the uimaj-example tutorials, I was able to create an
annotator that operates directly on a text document. Now I'm trying
to create a descriptor and annotator code (i.e. a 'primitive'
analysis engine) that simply looks at or accesses Sentence annotations.
In the descriptor (AnnotViewerDescr.xml), I've set the type system to
use the OpenNLPExampleTypes.xml (which specifies/references the
org.apache.uima.examples.opennlp.Sentence type). The capabillities
are set to take the Sentence type as input. This descriptor is
nearly identical to the OpenNLPTokenizer.xml file except that it
points to AnnotViewer.java and does not "output" any capabilities.
In the AnnotViewer.java file, I copied and pasted (from
Tokenizer.java) the AnnotationIndex declaration for the set of
sentences, and then print whether or not the sentence annotations
index contains anything. Here's the process method ...
<code>
public void process(JCas aJCas) throws AnalysisEngineProcessException {
AnnotationIndex sentenceIndex = aJCas.getAnnotationIndex(Sentence.type);
// iterate over Sentences
FSIterator sentenceIterator = sentenceIndex.iterator();
System.out.println(sentenceIterator.isValid());//.hasNext();
}
</code>
The sentenceIterator.isValid() is returning "false" when asked if it
has any sentence annotations. So the annotation is not happening or
is not being found. Any thoughts? What am I leaving out?
Many thanks,
Sam
