Porter stemmer: Python's NLTK vs. Java's OpenNLP: How to get *almost* identical Porter stemmers

README.md

Porter stemmer: Python's NLTK vs. Java's OpenNLP:

How to get almost equivalent Porter stemmers between Java and Python.

One remaining difference is how they handle non-dictionary words ending in "s":

Assuming you tokenize the following string after lowercase normalization using [a-z]\w+...

"EMI"s request for an appeal should be denied until a final judgment is entered. Document filed by Michael Robertson. (js) (Entered: 02/01/2012)"

NLTK will map token js -> j where OpenNLP is js -> js.

Note: There may be other differences yet to be discovered.

nltk-porter-stemmer.py

"""
Java's OpenNLP uses the original Porter stemmer algorithm.

NLTK implements Martin Porter's *revised* version of the stemmer.
You must set mode to 'ORIGINAL_ALGORITHM' to use the original.

See http://www.nltk.org/api/nltk.stem.html
"""

import nltk
porter = nltk.stem.porter.PorterStemmer(mode='ORIGINAL_ALGORITHM')

opennlp-porter-stemmer.java

import opennlp.tools.stemmer.PorterStemmer;
PorterStemmer stemmer = new PorterStemmer();