[
https://issues.apache.org/jira/browse/OPENNLP-316?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13122927#comment-13122927
]
Nicolas Hernandez commented on OPENNLP-316:
-------------------------------------------
Thanks Willian
looking with attention both at the link you sent and at some training data
available on the web site,
* I read "The O chunk tag is used for tokens which are not part of any
chunk." So the 'O' tag seems to be important.
* I observed on the training data that each sentence should be separated by
an empty line
I was not aware of this last point. I modified my data on purpose and magic...
the chunkerCrossValidator works...
I still do not know why it worked before with the chunkerTrainer and the
chunkerEvaluator...
I dared to try the same transformation with my training data for the Sentence
Detector... and It works... It works by adding an empty line between each
sentence...
I guess something must be fixed here.
The Trainer, the Evaluator and the CrossValidator should work with the same
format in input.
In my opinion, the newline as separator is unnecessary and this format
shouldn't be kept.
Best regards
> Evaluator and CrossValidator programs of the main analyzers throw exceptions
> ----------------------------------------------------------------------------
>
> Key: OPENNLP-316
> URL: https://issues.apache.org/jira/browse/OPENNLP-316
> Project: OpenNLP
> Issue Type: Bug
> Components: Chunker, POS Tagger, Sentence Detector, Tokenizer
> Affects Versions: tools-1.5.2-incubating
> Environment: Linux version 2.6.32-34-generic (buildd@yellow) (gcc
> version 4.4.3 (Ubuntu 4.4.3-4ubuntu5) ) #77-Ubuntu SMP Tue Sep 13 19:39:17
> UTC 2011
> java version "1.6.0_26"
> Java(TM) SE Runtime Environment (build 1.6.0_26-b03)
> Java HotSpot(TM) 64-Bit Server VM (build 20.1-b02, mixed mode)
> Reporter: Nicolas Hernandez
> Assignee: William Colen
> Fix For: tools-1.5.2-incubating
>
>
> Evaluator and CrossValidator programs of the main analyzers throw an
> exception when running
> (test performed on the 1.5.3 dist via command line)
> It seems that the SentenceDetector, Tokenizer, PosTagger and the
> chunker (at least) throw a java.lang.NullPointerException if the
> misclassified parameter is set to false or not present for the
> Evaluator programs.
> The Evaluator programs works (provide a result) when the
> misclassified parameter is set.
> The CrossValidator programs do not work at all.
> I have not test the other opennlp programs.
> See below some example of the runs.
> I tested on the examples from the documentation and also with my data.
> For the SentenceDetector I tested with with 1 000 and 1 000 000 sentences per
> line.
> Tell if you want more details or anything
> $opennlp SentenceDetectorEvaluator -encoding UTF-8 -model
> data/model/fr-sent.bin -data data/test/fr-sent.test
> Loading Sentence Detector model ... done (0,013s)
> Evaluating ... in thread "main" java.lang.NullPointerException
> at opennlp.tools.util.eval.Evaluator.evaluateSample(Evaluator.java:80)
> at opennlp.tools.util.eval.Evaluator.evaluate(Evaluator.java:98)
> at
> opennlp.tools.cmdline.sentdetect.SentenceDetectorEvaluatorTool.run(SentenceDetectorEvaluatorTool.java:80)
> at opennlp.tools.cmdline.CLI.main(CLI.java:191)
> $opennlp SentenceDetectorCrossValidator -encoding UTF-8 -lang fr -data
> data/train/fr-sent.train -misclassified true
> Indexing events using cutoff of 5
> Computing event counts... done. 0 events
> Indexing... done.
> Sorting and merging events... Done indexing.
> Incorporating indexed data for training...
> Exception in thread "main" java.lang.NullPointerException
> at opennlp.maxent.GISTrainer.trainModel(GISTrainer.java:263)
> at opennlp.maxent.GIS.trainModel(GIS.java:256)
> at opennlp.model.TrainUtil.train(TrainUtil.java:182)
> at
> opennlp.tools.sentdetect.SentenceDetectorME.train(SentenceDetectorME.java:283)
> at
> opennlp.tools.sentdetect.SDCrossValidator.evaluate(SDCrossValidator.java:104)
> at
> opennlp.tools.cmdline.sentdetect.SentenceDetectorCrossValidatorTool.run(SentenceDetectorCrossValidatorTool.java:98)
> at opennlp.tools.cmdline.CLI.main(CLI.java:191)
> $ opennlp TokenizerMEEvaluator -encoding UTF-8 -model
> data/model/fr-token.bin -data data/test/fr-token.test
> Loading Tokenizer model ... done (0,428s)
> Evaluating ... Exception in thread "main" java.lang.NullPointerException
> at opennlp.tools.util.eval.Evaluator.evaluateSample(Evaluator.java:76)
> at opennlp.tools.util.eval.Evaluator.evaluate(Evaluator.java:98)
> at
> opennlp.tools.cmdline.tokenizer.TokenizerMEEvaluatorTool.run(TokenizerMEEvaluatorTool.java:81)
> at opennlp.tools.cmdline.CLI.main(CLI.java:191)
> $ opennlp TokenizerCrossValidator -encoding UTF-8 -lang fr -data
> data/train/fr-token.train
> Indexing events using cutoff of 5
> Computing event counts... done. 100333 events
> Indexing... done.
> Sorting and merging events... done. Reduced 100333 events to 30168.
> Done indexing.
> Incorporating indexed data for training...
> done.
> Number of Event Tokens: 30168
> Number of Outcomes: 2
> Number of Predicates: 8287
> ...done.
> Computing model parameters ...
> Performing 100 iterations.
> 1: ... loglikelihood=-69545.53606709359 0.9337805108987073
> 2: ... loglikelihood=-18987.123809719425 0.9497872085953774
> ...
> 98: ... loglikelihood=-607.4216932752298 0.9989534848952987
> 99: ... loglikelihood=-603.2346954947699 0.9989734185163406
> 100: ... loglikelihood=-599.1235213848983 0.9989833853268616
> Exception in thread "main" java.lang.NullPointerException
> at opennlp.tools.util.eval.Evaluator.evaluateSample(Evaluator.java:76)
> at opennlp.tools.util.eval.Evaluator.evaluate(Evaluator.java:98)
> at
> opennlp.tools.tokenize.TokenizerCrossValidator.evaluate(TokenizerCrossValidator.java:98)
> at
> opennlp.tools.cmdline.tokenizer.TokenizerCrossValidatorTool.run(TokenizerCrossValidatorTool.java:94)
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
