Skip to content
Powerfull python wrapper for Stanford CoreNLP project
Branch: master
Clone or download
Sherin Thomas
Sherin Thomas Merge pull request #4 from matsair/patch-1
fixed typo in
Latest commit 821c03c Mar 21, 2017
Type Name Latest commit message Commit time
Failed to load latest commit information.
corenlp_pywrap added function for tregex (tree node matching) Jan 2, 2017
LICENCE.txt added files and changed directory structure for Jul 31, 2016 Update Mar 21, 2017
setup.cfg version 1.0.5, changed fillers from null to 0 Aug 14, 2016

corenlp_pywrap 1.0.5

Powerfull python wrapper for Stanford CoreNLP project

GitHub release license PyPI

CoreNLP v3.6.0

  • Update your version (both CoreNLP and corenlp_pywrap) for bug fixes and more features
  • Works only with python 3.x

Pywrap Doc


pip install corenlp_pywrap


pip3 install corenlp_pywrap


from corenlp_pywrap import pywrap
cn = pywrap.CoreNLP(url='http://localhost:9000', annotator_list=full_annotator_list)
#full_annotator_list = ["tokenize", "cleanxml", "ssplit", "pos", "lemma", "ner", "regexner", "truecase", "parse", "depparse", "dcoref", "relation", "natlog", "quote"]

#Calling basic function which would return a 'requests' object
out = cn.basic(data, out_format='json')

Remember 'out' would be 'requests' object, you can get information by using out.text or out.json()

Pywrap does not inherently support 'Sentiment' now as the downloadable server version of CoreNLP doesn't have 'Sentiment' support. But there is a hack for you to use (if you are sure that your server version is the newest one and has the support)

  • You can give sentiment as an attribute to annotator_list while instantiating the class object
  • or
annotator_list = CoreNLP.full_annotator_list + ['sentiment']

Custom Function

  • arrange() can be used for getting formatted output
  • Format is given below
token_dict = {
  • arrange() would return token_dict which is in the above format.
  • arrange() would work only for 'sentences' now. More features like 'enhanceddependancies' are coming in future releases
  • Usage:
from corenlp_pywrap import pywrap
cn = pywrap.CoreNLP(url='http://localhost:9000', annotator_list=full_annotator_list)
#full_annotator_list = ["tokenize", "cleanxml", "ssplit", "pos", "lemma", "ner", "regexner", "truecase", "parse", "depparse", "dcoref", "relation", "natlog", "quote"]

#custom function
token_dict = cn.arrange(data)

#example out: token_dict['index'] would give you something like this - [1,2,3,4]

Server Instantiation Error

  • If you don't have the CoreNLP server downloaded, please download the server here
  • Make sure you have Jave 8+ version installed
  • CD to the downloaded folder
  • Follow below commands
# Run the server using all jars in the current directory (e.g., the CoreNLP home directory)
# port and timeout are optional
java -mx4g -cp "*" edu.stanford.nlp.pipeline.StanfordCoreNLPServer [port] [timeout]
  • Verify the server instance in the browser


replace 'port' with the port number you have given. if you didn't give any port number, port would be 9000


Debugging & Logging

  • Pywrap using logging module for logging and debugging.

  • Default logging level is set to 'warning'

  • If you need more verbose logs for debugging or logging purpose make changes to the logging values

  • Default log facilities

root = logging.getLogger('Root')

lhandler = logging.StreamHandler(sys.stdout)
formatter = logging.Formatter(
                '%(asctime)s [%(name)s]:%(levelname)s - %(message)s',
                '%Y-%m-%d %H:%M:%S')
  • You can modify each of them just like below given example
import corenlp_pywrap as cp
import logging
You can’t perform that action at this time.