This is a Python wrapper for Stanford University's NLP group's Java-based CoreNLP tools. It can either be imported as a module or run as a JSON-RPC server. Because it uses many large trained models (requiring 3GB RAM on 64-bit machines and usually a few minutes loading time), most applications will probably want to run it as a server.
- Python interface to Stanford CoreNLP tools: tagging, phrase-structure parsing, dependency parsing, named-entity recognition, and coreference resolution.
- Runs an JSON-RPC server that wraps the Java server and outputs JSON.
- Outputs parse trees which can be used by nltk.
It depends on pexpect and includes and uses code from jsonrpc and python-progressbar.
It runs the Stanford CoreNLP jar in a separate process, communicates with the java process using its command-line interface, and makes assumptions about the output of the parser in order to parse it into a Python dict object and transfer it using JSON. The parser will break if the output changes significantly, but it has been tested on Core NLP tools version 3.5.2 released 2015-04-20.
To use this program you must download and unpack the compressed file containing Stanford's CoreNLP package. By default, corenlp.py
looks for the Stanford Core NLP folder as a subdirectory of where the script is being run. In other words:
sudo pip install pexpect unidecode
git clone git://github.com/dasmith/stanford-corenlp-python.git
cd stanford-corenlp-python
wget http://nlp.stanford.edu/software/stanford-corenlp-full-2015-04-20.zip
unzip stanford-corenlp-full-2015-04-20.zip
Then launch the server:
python corenlp.py
Optionally, you can specify a host or port:
python corenlp.py -H 0.0.0.0 -p 3456
That will run a public JSON-RPC server on port 3456.
Assuming you are running on port 8080, the code in client.py
shows an example parse:
import jsonrpc
from simplejson import loads
server = jsonrpc.ServerProxy(jsonrpc.JsonRpc20(),
jsonrpc.TransportTcpIp(addr=("127.0.0.1", 8080)))
result = loads(server.parse("Hello world. It is so beautiful"))
print "Result", result
That returns a dictionary containing the keys sentences
and coref
. The key sentences
contains a list of dictionaries for each sentence, which contain parsetree
, text
, tuples
containing the dependencies, and words
, containing information about parts of speech, recognized named-entities, etc:
{u'sentences': [{u'parsetree': u'(ROOT (S (VP (NP (INTJ (UH Hello)) (NP (NN world)))) (. !)))',
u'text': u'Hello world!',
u'tuples': [[u'dep', u'world', u'Hello'],
[u'root', u'ROOT', u'world']],
u'words': [[u'Hello',
{u'CharacterOffsetBegin': u'0',
u'CharacterOffsetEnd': u'5',
u'Lemma': u'hello',
u'NamedEntityTag': u'O',
u'PartOfSpeech': u'UH'}],
[u'world',
{u'CharacterOffsetBegin': u'6',
u'CharacterOffsetEnd': u'11',
u'Lemma': u'world',
u'NamedEntityTag': u'O',
u'PartOfSpeech': u'NN'}],
[u'!',
{u'CharacterOffsetBegin': u'11',
u'CharacterOffsetEnd': u'12',
u'Lemma': u'!',
u'NamedEntityTag': u'O',
u'PartOfSpeech': u'.'}]]},
{u'parsetree': u'(ROOT (S (NP (PRP It)) (VP (VBZ is) (ADJP (RB so) (JJ beautiful))) (. .)))',
u'text': u'It is so beautiful.',
u'tuples': [[u'nsubj', u'beautiful', u'It'],
[u'cop', u'beautiful', u'is'],
[u'advmod', u'beautiful', u'so'],
[u'root', u'ROOT', u'beautiful']],
u'words': [[u'It',
{u'CharacterOffsetBegin': u'14',
u'CharacterOffsetEnd': u'16',
u'Lemma': u'it',
u'NamedEntityTag': u'O',
u'PartOfSpeech': u'PRP'}],
[u'is',
{u'CharacterOffsetBegin': u'17',
u'CharacterOffsetEnd': u'19',
u'Lemma': u'be',
u'NamedEntityTag': u'O',
u'PartOfSpeech': u'VBZ'}],
[u'so',
{u'CharacterOffsetBegin': u'20',
u'CharacterOffsetEnd': u'22',
u'Lemma': u'so',
u'NamedEntityTag': u'O',
u'PartOfSpeech': u'RB'}],
[u'beautiful',
{u'CharacterOffsetBegin': u'23',
u'CharacterOffsetEnd': u'32',
u'Lemma': u'beautiful',
u'NamedEntityTag': u'O',
u'PartOfSpeech': u'JJ'}],
[u'.',
{u'CharacterOffsetBegin': u'32',
u'CharacterOffsetEnd': u'33',
u'Lemma': u'.',
u'NamedEntityTag': u'O',
u'PartOfSpeech': u'.'}]]}],
u'coref': [[[[u'It', 1, 0, 0, 1], [u'Hello world', 0, 1, 0, 2]]]]}
To use it in a regular script (useful for debugging), load the module instead:
from corenlp import *
corenlp = StanfordCoreNLP() # wait a few minutes...
corenlp.parse("Parse this sentence.")
The server, StanfordCoreNLP()
, takes an optional argument corenlp_path
which specifies the path to the jar files. The default value is StanfordCoreNLP(corenlp_path="./stanford-corenlp-full-2015-04-20/")
.
The library supports coreference resolution, which means pronouns can be "dereferenced." If an entry in the coref
list is, [u'Hello world', 0, 1, 0, 2]
, the numbers mean:
- 0 = The reference appears in the 0th sentence (e.g. "Hello world")
- 1 = The 2nd token, "world", is the headword of that sentence
- 0 = 'Hello world' begins at the 0th token in the sentence
- 2 = 'Hello world' ends before the 2nd token in the sentence.
Stanford CoreNLP tools require a large amount of free memory. Java 5+ uses about 50% more RAM on 64-bit machines than 32-bit machines. 32-bit machine users can lower the memory requirements by changing -Xmx3g
to -Xmx2g
or even less.
If pexpect timesout while loading models, check to make sure you have enough memory and can run the server alone without your kernel killing the java process:
java -cp stanford-corenlp-2015-04-20.jar:stanford-corenlp-3.4.1-models.jar:xom.jar:joda-time.jar -Xmx3g edu.stanford.nlp.pipeline.StanfordCoreNLP -props default.properties
You can reach me, Dustin Smith, by sending a message on GitHub or through email (contact information is available on my webpage).
This is free and open source software and has benefited from the contribution and feedback of others. Like Stanford's CoreNLP tools, it is covered under the GNU General Public License v2 +, which in short means that modifications to this program must maintain the same free and open source distribution policy.
I gratefully welcome bug fixes and new features. If you have forked this repository, please submit a pull request so others can benefit from your contributions. This project has already benefited from contributions from these members of the open source community:
- Emilio Monti
- Justin Cheng
- Abhaya Agarwal
Thank you!
Maintainers of the Core NLP library at Stanford keep an updated list of wrappers and extensions. See Brendan O'Connor's stanford_corenlp_pywrapper for a different approach more suited to batch processing.