Spacy wrapper for OpenNLP models

This package provides a very simple wrapper for the OpenNLP models for spaCy which is working through command line calls to the OpenNLP tools. It is not very efficient, so the main purpose check and test existing OpenNLP models.

Installation

Install opennlp with apt-get or brew:

brew install opennlp # on Mac
apt-get install opennlp # on Linux

For Windows, you can download the binaries from here. OpenNLP requires Java, so make sure you have Java installed. Then add the path to the OpenNLP binaries to your PATH environment variable. Check if OpenNLP is installed correctly by running the following command:

opennlp

Install the package with pip:

pip install spacy-opennlp

Usage

The usage is very simple. You can load the models with the load function, where path is the path to the OpenNLP models (expected 4 models, e.g. for Hebrew: he-lemmatizer.bin, he-pos.bin, he-sent.bin, he-token.bin), and lang is the language code (e.g. he for Hebrew, so package will look for {lang}-lemmatizer.bin, {lang}-pos.bin, {lang}-sent.bin, {lang}-token.bin in the path directory).

import spacy_opennlp


nlp = spacy_opennlp.load(
    "he",
    path="<PATH TO MODELS THERE>",
)
text = "בתל אביב יש כמה רכבת קלה."
doc = nlp(text)
for sentence in doc.sents:
    print(sentence.text)
    
    for token in sentence:
        print(token.text, token.lemma_, token._.opennlp_tag)
    print()