Hello,
started to look into the packaging issue, but I cannot build it cause
of failing tests. The test should run trough before we release it.
It looks like the test depend on our English models, I can for example
get this exception:
java.io.FileNotFoundException:
/home/xyz/opennlp/sandbox/opennlp-similarity/src/test/resources/models/en-chunker.bin
(No such file or directory)
at java.io.FileInputStream.open(Native Method)
at java.io.FileInputStream.<init>(FileInputStream.java:137)
at java.io.FileInputStream.<init>(FileInputStream.java:96)
at
opennlp.tools.textsimilarity.chunker2matcher.ParserChunker2MatcherProcessor.initializeChunker(ParserChunker2MatcherProcessor.java:658)
at
opennlp.tools.textsimilarity.chunker2matcher.ParserChunker2MatcherProcessor.<init>(ParserChunker2MatcherProcessor.java:119)
at
opennlp.tools.textsimilarity.chunker2matcher.ParserChunker2MatcherProcessor.getInstance(ParserChunker2MatcherProcessor.java:140)
at
opennlp.tools.textsimilarity.chunker2matcher.ParserChunker2MatcherProcessorTest.testGroupedPhrasesFormer(ParserChunker2MatcherProcessorTest.java:34)
The models cannot be included in subversion or in the distribution
because they are not
AL 2.0 licensed. So we should not have tests depending on them.
Can you fix this? In my opinion the tests should just use pre-annotated
data and do not use
any of our models to produce this data on the fly.
Please get a fresh checkout of the opennlp-similarity folder, and try to
get it build
with mvn install. That needs to work out of the box. There were also
other test failures,
e.g. one where it could not parse a sentence.
Jörn
On 09/19/2012 11:57 PM, Boris Galitsky wrote:
Hi guys
I would like to complete this project.This tickets
https://issues.apache.org/jira/browse/OPENNLP-497
is supposed to conclude it, I edited pom.xml and added files similar to what
apache-opennlp-1.5.2-incubating-bin.zip has.
I manually formed the distribution archive and attached it to this ticket.
As an additional round of verification that Similarity is working, besides a
number of application areas (which live in junits) I also built a SOLR search
handler which re-rankssearch results according to similarity of parse trees for
questions and answers. This way SOLR uses OpenNLP to improve search relevance
(will create a ticket and add the request handler).
What are the next steps?
Best regardsBoris
