问题
I use the following command to serve a corenlp server for German language models which are downloaded as jar in the classpath , but it does not output german tags or parse but loads only english models:
java -mx4g -cp "*" edu.stanford.nlp.pipeline.StanfordCoreNLPServer -props ./german.prop
german.prop contents:
annotators = tokenize, ssplit, pos, depparse, parse
tokenize.language = de
pos.model = edu/stanford/nlp/models/pos-tagger/german/german-hgc.tagger
ner.model = edu/stanford/nlp/models/ner/german.hgc_175m_600.crf.ser.gz
ner.applyNumericClassifiers = false
ner.useSUTime = false
parse.model = edu/stanford/nlp/models/lexparser/germanFactored.ser.gz
depparse.model = edu/stanford/nlp/models/parser/nndep/UD_German.gz
client command:
wget --post-data ' Meine Mutter ist aus Wuppertal' 'localhost:9000/?properties"="{"tokenize.whitespace":"true","annotators":"tokenize, ssplit, pos, depparse, parse","outputFormat":"text","tokenize.language" :"de" ,
"pos.model":" edu/stanford/nlp/models/pos-tagger/german/german-hgc.tagger",
"depparse.model" : "edu/stanford/nlp/models/parser/nndep/UD_German.gz",
"parse.model" : "edu/stanford/nlp/models/lexparser/germanFactored.ser.gz"
}' -O -
I get following incorrect output:
{"dep":"dep","governor":4,"governorGloss":"aus","dependent":5,"dependentGloss":"Wuppertal"}],"openie":[{"subject":"Wuppertal","subjectSpan":[4,5],"relation":"is ist aus of","relationSpan":[2,4],"object":"Meine Mutter","objectSpan":[0,2]}],"tokens":[{"index":1,"word":"Meine","originalText":"Meine","lemma":"Meine","characterOffsetBegin":1,"characterOffsetEnd":6,"pos":"NNP","ner":"PERSON","speaker":"PER0","before":" ","after":" "},{"index":2,"word":"Mutter","originalText":"Mutter","lemma":"Mutter","characterOffsetBegin":7,"characterOffsetEnd":13,"pos":"NNP","ner":"PERSON","speaker":"PER0","before":" ","after":" "},{"index":3,"word":"ist","originalText":"ist","lemma":"ist","characterOffsetBegin":14,"characterOffsetEnd":17,"pos":"NN","ner":"O","speaker":"PER0","before":" ","after":" "},{"index":4,"word":"aus","originalText":"aus","lemma":"aus","characterOffsetBegin":18,"characterOffsetEnd":21,"pos":"NN","ner":"O","speaker":"PER0","before":" ","after":" "},{"index":5,"word":"Wuppertal","originalText":"Wuppertal","lemma":"Wuppertal","characterOffsetBegin":22,"characterOffsetEnd":31,"pos":"NNP","ner":"LOCATI100%[==========================================================================>] 2,
in the server log I see it loads english models eventhough it lists german models on startup:
pos.model=edu/stanford/nlp/models/pos-tagger/ge...
parse.model=edu/stanford/nlp/models/lexparser/ger...
tokenize.language=de
depparse.model=edu/stanford/nlp/models/parser/nndep/...
annotators=tokenize, ssplit, pos, depparse, parse
Starting server on port 9000 with timeout of 5000 milliseconds.
StanfordCoreNLPServer listening at /0:0:0:0:0:0:0:0:9000
[/203.:61563] API call w/annotators tokenize,ssplit,pos,depparse
Die Katze liegt auf der Matte.
[pool-1-thread-1] INFO edu.stanford.nlp.pipeline.StanfordCoreNLP - Adding annotator tokenize
[pool-1-thread-1] INFO edu.stanford.nlp.pipeline.TokenizerAnnotator - TokenizerAnnotator: No tokenizer type provided. Defaulting to PTBTokenizer.
[pool-1-thread-1] INFO edu.stanford.nlp.pipeline.StanfordCoreNLP - Adding annotator ssplit
[pool-1-thread-1] INFO edu.stanford.nlp.pipeline.StanfordCoreNLP - Adding annotator pos
Reading POS tagger model from edu/stanford/nlp/models/pos-tagger/english-left3words/english-left3words-distsim.tagger ... done [1.5 sec].
[pool-1-thread-1] INFO edu.stanford.nlp.pipeline.StanfordCoreNLP - Adding annotator depparse
Loading depparse model file: edu/stanford/nlp/models/parser/nndep/english_UD.gz ...
PreComputed 100000, Elapsed Time: 1.396 (s)
The following question for same error in french models also points to the same problem but even after following , it does not resolve the problem for the server case, I am able to get the correct output without using the server and just using the edu.stanford.nlp.pipeline.StanfordCoreNLP command , it is the server command edu.stanford.nlp.pipeline.StanfordCoreNLPServer which defaults to english:
French dependency parsing using CoreNLP
回答1:
There have been some issues with getting the foreign language stuff to work on the server.
If you use the latest release available at our GitHub site, it should work.
The GitHub site is here: https://github.com/stanfordnlp/CoreNLP
That link has instructions for building a jar with the latest version of the code.
I ran this command on some sample German text and it looks like it works fine:
wget --post-data '<sample german text>' 'localhost:9000/?properties={"pipelineLanguage":"german","annotators":"tokenize,ssplit,pos,ner,parse", "parse.model":"edu/stanford/nlp/models/lexparser/germanFactored.ser.gz","tokenize.language":"de","pos.model":"edu/stanford/nlp/models/pos-tagger/german/german-hgc.tagger", "ner.model":"edu/stanford/nlp/models/ner/german.hgc_175m_600.crf.ser.gz", "ner.applyNumericClassifiers":"false", "ner.useSUTime":"false"}' -O -
I should note that the neural net German dependency parser is completely broken and we are working on fixing it soon, so you should just use the German settings I specified in that command.
More info on the server can be found here: http://stanfordnlp.github.io/CoreNLP/corenlp-server.html
来源：https://stackoverflow.com/questions/39688652/german-corenlp-model-defaulting-to-english-models