Skip to content
Match tokenized words and phrases within the original, untokenized, often messy, text.
Python Shell
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
doc
match
.travis.yml
LICENSE
README.md
README.txt
build.sh
meta.yaml
setup.py
test_match.py

README.md

match

Build Status

Binstar BadgeBinstar Badge

Latest VersionDownloads

The purpose of the module Match is to get the offsets (as well as the string between those offsets, for debugging) of a cleaned-up, tokenized string from its original, untokenized source. "Big deal," you might say, but this is actually a pretty difficult task if the original text is sufficiently messy, not to mention rife with Unicode characters.

Consider some text, stored in a variable original_text, like:

I   am writing a letter !  Sometimes,I forget to put spaces (and do weird stuff with punctuation)  ?  J'aurai une pomme, s'il vous plâit !

This will/should/might be properly tokenized as:

[[u'I', u'am', u'writing', u'a', u'letter', u'!'],
 [u'Sometimes', u',', u'I', u'forget', u'to', u'put', u'spaces', u'-LRB-', u'and', u'do', u'weird', u'stuff', u'with', u'punctuation', u'-RRB-', u'?'],
 [u"J'aurai", u'une', u'pomme', u',', u"s'il", u'vous', u'pl\xe2it', u'!']]

Now:

In [22]: Match.match(original_text, [u'-LRB-', u'and', u'do', u'weird', u'stuff', u'with', u'punctuation', u'-RRB-'])
Out[22]: [(60, 97, u'(and do weird stuff with punctuation)')]

In [23]: Match.match(original_text, [u'I', u'am', u'writing', u'a', u'letter', u'!'])
Out[23]: [(0, 25, u'I   am writing a letter !')]

In [24]: Match.match(original_text, [u"s'il", u'vous', u'pl\xe2it', u'!'])
Out[24]: [(121, 138, u"s'il vous pl\xe2it !")]

The return type from match() is a list because it will return all occurrences of the argument, be it a list of tokens or a single string (word):

In [25]: Match.match(original_text, "I")
Out[25]: [(0, 1, u'I'), (37, 38, u'I')]

When passing in a single string, match() is expecting that string to be a single word or token. Thus, the following does not work as one would expect:

In [5]: Match.match("****because,the****", "because , the")
Out[5]: []

Try "because , the".split(" ") instead.

For convenience, a function called match_lines() is provided:

In [26]: Match.match_lines(original_text, [[u'-LRB-', u'and', u'do', u'weird', u'stuff', u'with', u'punctuation', u'-RRB-'], [u'I', u'am', u'writing', u'a', u'letter', u'!'], "I"])
Out[26]: 
[(0, 1, u'I'),
 (0, 25, u'I   am writing a letter !'),
 (37, 38, u'I'),
 (60, 97, u'(and do weird stuff with punctuation)')]

The values returned will always be sorted by their offsets.

Installation

pip install match, or for Mac OS X and 64-bit Linux:

$ conda install -c https://conda.binstar.org/dmnapolitano match

Requirements

  • Python >= 2.7 (works with both 2.7 and 3.4!)
  • nltk
  • regex

Documentation

Viewable here.

You can’t perform that action at this time.