NLTK interface with Apache OpenNLP
Copyright (C) 2017 Paulius Danenas
- Apache OpenNLP
- Python 2.7 or Python 3
- NLTK
Tested with OpenNLP 1.8 (using models built with 1.5), Python 2.7/3.5/3.6 and NLTK 3.5
Before you install the nltk-opennlp package please ensure you
have downloaded and installed the Apache OpenNLP
itself. You will also need different tagger/chunker models; some of them are provided in
this repository
OpenNLP setup can be automated using build.py script which will automatically download OpenNLP binaries and models for predefined languages.
By default, if they will be installed into current directory. pybuilder package is required to run this script; it can be installed with pip:
pip3 install pybuilder
After cloning this repo, run pyb in its directory which contains the build.py file. Verify if the installation was successful by running tests in tests.py
Tagging a sentence from Python:
from nltk_opennlp.taggers import OpenNLPTagger
tt = OpenNLPTagger(language='en',
path_to_bin=os.path.join('/path/to/opennlp/installation', 'bin'),
path_to_model=os.path.join('/path/to/opennlp/models', 'en-pos-maxent.bin'))
phrase = 'Pierre Vinken , 61 years old , will join the board as a nonexecutive director Nov. 29 .'
sentence = tt.tag(phrase)The output is a list of (token, tag):
[('Pierre', 'NNP'), ('Vinken', 'NNP'), (',', ','), ('61', 'CD'),
('years', 'NNS'), ('old', 'JJ'), (',', ','), ('will', 'MD'),
('join', 'VB'), ('the', 'DT'), ('board', 'NN'), ('as', 'IN'),
('a', 'DT'), ('nonexecutive', 'JJ'), ('director', 'NN'),
('Nov.', 'NNP'), ('29', 'CD'), ('.', '.')]
Chunking the same sentence from Python will produce a parse tree:
from nltk_opennlp.chunkers import OpenNLPChunker
from nltk_opennlp.taggers import OpenNLPTagger
tt = OpenNLPTagger(language='en',
path_to_bin=os.path.join('/path/to/opennlp/installation', 'bin'),
path_to_model=os.path.join('/path/to/opennlp/models', 'en-pos-maxent.bin'))
phrase = 'Pierre Vinken , 61 years old , will join the board as a nonexecutive director Nov. 29 .'
sentence = tt.tag(phrase)
cp = OpenNLPChunker(path_to_bin=os.path.join('/path/to/opennlp/installation', 'bin'),
path_to_chunker=os.path.join('/path/to/opennlp/models', 'en-chunker.bin'))
print(cp.parse(sentence))The output is a parse tree:
(S (NP Pierre/NNP Vinken/NNP) ,/, (NP 61/CD years/NNS) (ADJP old/JJ) ,/, (VP will/MD join/VB) (NP the/DT board/NN) (PP as/IN) (NP a/DT nonexecutive/JJ director/NN) (NP Nov./NNP 29/CD) ./.)
Note, that is possible to use PUNC tag to tag standalone punctuation marks, using use_punc_tag parameter. After setting this param, the output would be come as following:
(S (PERSON Pierre/NNP Vinken/NNP) ,/PUNC (NP 61/CD years/NNS) (ADJP old/JJ) ,/PUNC (VP will/MD join/VB) (PERSON Martin/NNP Vinken/NNP) (PP as/IN) (NP a/DT nonexecutive/JJ director/NN) (NP Nov./NNP 29/CD) ./PUNC)
Tagging a german sentence from Python is similar, just need to use diferent language and pre-trained model:
from nltk_opennlp.taggers import OpenNLPTagger
tt = OpenNLPTagger(language='de',
path_to_bin=os.path.join('/path/to/opennlp/installation', 'bin'),
path_to_model=os.path.join('/path/to/opennlp/models', 'de-pos-maxent.bin'))
tt.tag('Das Haus hat einen großen hübschen Garten.')The output is a list of (token, tag):
[('Das', 'ART'), ('Haus', 'NN'), ('hat', 'VAFIN'), ('einen', 'ART'), (
'großen', 'ADJA'), ('hübcbschen', 'ADJA'), ('Garten.', 'NN')]
This module also supports named entity recognition, which allows to tag particular types of entities. Again, chunking
is performed on the set of (token, tag) entries (note, that NLTK taggers could be used instead of OpenNLPTagger):
from nltk_opennlp.chunkers import OpenNERChunker
language='en'
tt = OpenNLPTagger(language=language,
path_to_bin=os.path.join(opennlp_dir, 'apache-opennlp', 'bin'),
path_to_model=os.path.join(opennlp_dir, 'opennlp_models', 'en-pos-maxent.bin'))
phrase = 'Pierre Vinken , 61 years old , will join Martin Vinken as a nonexecutive director Nov. 29 .'
sentence = tt.tag(phrase)
cp = OpenNERChunker(path_to_bin=os.path.join(opennlp_dir, 'apache-opennlp', 'bin'),
path_to_chunker=os.path.join(opennlp_dir, 'opennlp_models', '{}-chunker.bin'.format(language)),
path_to_ner_model=os.path.join(opennlp_dir, 'opennlp_models', '{}-ner-person.bin'.format(language)),
use_punc_tag=True)
print(cp.parse(sentence))The output is a chunk parse tree with particular types of entities:
(S (PERSON Pierre/NNP Vinken/NNP) ,/, (NP 61/CD years/NNS) (ADJP old/JJ) ,/, (VP will/MD join/VB) (PERSON Martin/NNP Vinken/NNP) (PP as/IN) (NP a/DT nonexecutive/JJ director/NN) (NP Nov./NNP 29/CD) ./.)
A multi-tagger option is similar, except that it allows to set multiple NER models for tagging:
from nltk_opennlp.chunkers import OpenNERChunkerMulti
language='en'
phrase = 'John Haddock , 32 years old male , travelled to Cambridge , USA in October 20 while paying 6.50 dollars for the ticket'
sentence = tt.tag(phrase)
cp = OpenNERChunkerMulti(language=language,
path_to_bin=os.path.join(opennlp_dir, 'apache-opennlp', 'bin'),
path_to_chunker=os.path.join(opennlp_dir, 'opennlp_models', '{}-chunker.bin'.format(language)),
ner_models=[os.path.join(opennlp_dir, 'opennlp_models', '{}-ner-person.bin'.format(language)),
os.path.join(opennlp_dir, 'opennlp_models', '{}-ner-date.bin'.format(language)),
os.path.join(opennlp_dir, 'opennlp_models', '{}-ner-location.bin'.format(language)),
os.path.join(opennlp_dir, 'opennlp_models', '{}-ner-time.bin'.format(language)),
os.path.join(opennlp_dir, 'opennlp_models', '{}-ner-money.bin'.format(language))])
print(cp.parse(sentence))The resuting chunk tree contains multiple types of identified entities:
(S (PERSON John/NNP Haddock/NNP) ,/, (NP 32/CD years/NNS) (NP old/JJ male/NN) ,/, (VP travelled/VBN) (PP to/TO) (LOCATION Cambridge/NNP) ,/, (NP USA/NNP) (PP in/IN) (DATE October/NNP 20/CD) (PP while/IN) (VP paying/VBG) (NP 6.50/CD dollars/NNS) (PP for/IN) (NP the/DT ticket/NN))