Skip to main content
Help us improve Python packaging – donate today!

English word segmentation.

Project Description

WordSegment is an Apache2 licensed module for English word segmentation, written in pure-Python, and based on a trillion-word corpus.

Based on code from the chapter “Natural Language Corpus Data” by Peter Norvig from the book “Beautiful Data” (Segaran and Hammerbacher, 2009).

Data files are derived from the Google Web Trillion Word Corpus, as described by Thorsten Brants and Alex Franz, and distributed by the Linguistic Data Consortium. This module contains only a subset of that data. The unigram data includes only the most common 333,000 words. Similarly, bigram data includes only the most common 250,000 phrases. Every word and phrase is lowercased with punctuation removed.

Features

  • Pure-Python
  • Fully documented
  • 100% Test Coverage
  • Includes unigram and bigram data
  • Command line interface for batch processing
  • Easy to hack (e.g. different scoring, new data, different language)
  • Developed on Python 2.7
  • Tested on CPython 2.6, 2.7, 3.2, 3.3, 3.4, 3.5, 3.6 and PyPy, PyPy3
  • Tested on Windows, Mac OS X, and Linux
  • Tested using Travis CI and AppVeyor CI
https://ci.appveyor.com/api/projects/status/github/grantjenks/python-wordsegment?branch=master&svg=true

Quickstart

Installing WordSegment is simple with pip:

$ pip install wordsegment

You can access documentation in the interpreter with Python’s built-in help function:

>>> import wordsegment
>>> help(wordsegment)

Tutorial

In your own Python programs, you’ll mostly want to use segment to divide a phrase into a list of its parts:

>>> from wordsegment import load, segment
>>> load()
>>> segment('thisisatest')
['this', 'is', 'a', 'test']

The load function reads and parses the unigrams and bigrams data from disk. Loading the data only needs to be done once.

WordSegment also provides a command-line interface for batch processing. This interface accepts two arguments: in-file and out-file. Lines from in-file are iteratively segmented, joined by a space, and written to out-file. Input and output default to stdin and stdout respectively.

$ echo thisisatest | python -m wordsegment
this is a test

If you want to run WordSegment as a kind of server process then use Python’s -u option for unbuffered output. You can also set PYTHONUNBUFFERED=1 in the environment.

>>> import subprocess as sp
>>> wordsegment = sp.Popen(
        ['python', '-um', 'wordsegment'],
        stdin=sp.PIPE, stdout=sp.PIPE, stderr=sp.STDOUT)
>>> wordsegment.stdin.write('thisisatest\n')
>>> wordsegment.stdout.readline()
'this is a test\n'
>>> wordsegment.stdin.write('workswithotherlanguages\n')
>>> wordsegment.stdout.readline()
'works with other languages\n'
>>> wordsegment.stdin.close()
>>> wordsegment.wait()  # Process exit code.
0

The maximum segmented word length is 24 characters. Neither the unigram nor bigram data contain words exceeding that length. The corpus also excludes punctuation and all letters have been lowercased. Before segmenting text, clean is called to transform the input to a canonical form:

>>> from wordsegment import clean
>>> clean('She said, "Python rocks!"')
'shesaidpythonrocks'
>>> segment('She said, "Python rocks!"')
['she', 'said', 'python', 'rocks']

Sometimes its interesting to explore the unigram and bigram counts themselves. These are stored in Python dictionaries mapping word to count.

>>> import wordsegment as ws
>>> ws.load()
>>> ws.UNIGRAMS['the']
23135851162.0
>>> ws.UNIGRAMS['gray']
21424658.0
>>> ws.UNIGRAMS['grey']
18276942.0

Above we see that the spelling gray is more common than the spelling grey.

Bigrams are joined by a space:

>>> import heapq
>>> from pprint import pprint
>>> from operator import itemgetter
>>> pprint(heapq.nlargest(10, ws.BIGRAMS.items(), itemgetter(1)))
[('of the', 2766332391.0),
 ('in the', 1628795324.0),
 ('to the', 1139248999.0),
 ('on the', 800328815.0),
 ('for the', 692874802.0),
 ('and the', 629726893.0),
 ('to be', 505148997.0),
 ('is a', 476718990.0),
 ('with the', 461331348.0),
 ('from the', 428303219.0)]

Some bigrams begin with <s>. This is to indicate the start of a bigram:

>>> ws.BIGRAMS['<s> where']
15419048.0
>>> ws.BIGRAMS['<s> what']
11779290.0

The unigrams and bigrams data is stored in the wordsegment directory in the unigrams.txt and bigrams.txt files respectively.

WordSegment License

Copyright 2018 Grant Jenks

Licensed under the Apache License, Version 2.0 (the “License”); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an “AS IS” BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

Release history Release notifications

This version
History Node

1.2.0

History Node

1.1.6

History Node

1.1.5

History Node

1.1.4

History Node

1.1.3

History Node

0.8.0

History Node

0.7.3

History Node

0.7.2

History Node

0.7.1

History Node

0.6.2

History Node

0.6.1

History Node

0.5.3

History Node

0.5.2

History Node

0.3

History Node

0.2

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Filename, size & hash SHA256 hash help File type Python version Upload date
wordsegment-1.2.0-py2.py3-none-any.whl (4.4 MB) Copy SHA256 hash SHA256 Wheel py2.py3 Mar 1, 2018
wordsegment-1.2.0.tar.gz (4.4 MB) Copy SHA256 hash SHA256 Source None Mar 1, 2018

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging CloudAMQP CloudAMQP RabbitMQ AWS AWS Cloud computing Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page