NER is over writing the custom NERin stanford NLP

问题

In the stanford nlp, I used a pattern to match the phone number in regexner. But the NER is over writing it as Number.

If I remove the ner annotation then it is showing as PHONE_NUMBER. Can any one of you please help me.

Thanks in Advance.

Here is my regexner line:

^(?:(?:\+|0{0,2})91(\s*[\-]\s*)?|[0]?)?[789]\d{9}$  PHONENUMBER

回答1:

java command:

java -Xmx10g edu.stanford.nlp.pipeline.StanfordCoreNLP -annotators tokenize,ssplit,pos,lemma,ner -file phone-number-example.txt -outputFormat text -ner.fine.regexner.mapping phone-number-regex.rules

example text:

I will call him at 555-555-5555

format of rules file:

555-555-5555    PHONE_NUMBER    NUMBER  1

(note the columns are tab delimited)

The fine-grained NER will be applied after the statistical NER. You can also build a custom regexner and run it after the statistical model. The key is telling it to overwrite the NUMBER tag (which is indicated in the third column).

回答2:

^(?:(?:\+|0{0,2})91(\s*[\-]\s*)?|[0]?)?[789]\d{9}$  PHONENUMBER NUMBER

this worked the column after the CUSTOM NER column to overwrite

来源：https://stackoverflow.com/questions/50678418/ner-is-over-writing-the-custom-nerin-stanford-nlp