Greeb [grʲip] is a simple Unicode-aware text segmentator based on regular expressions. The API documentation is available on RubyDoc.info.
Add this line to your application's Gemfile:
gem 'greeb'
And then execute:
$ bundle
Or install it yourself as:
$ gem install greeb
Greeb can approach such essential text processing problems as tokenization and segmentation. There are two ways to use it: (1) as a command-line application, (2) as a Ruby library.
The greeb
application reads the input text from STDIN
and
writes one token per line to STDOUT
.
% echo 'Hello http://nlpub.ru guys, how are you?' | greeb
Hello
http://nlpub.ru
guys
,
how
are
you
?
Greeb has a very convenient API that makes you happy.
pp Greeb::Tokenizer.tokenize('Hello!')
=begin
[#<struct Greeb::Span from=0, to=5, type=:letter>,
#<struct Greeb::Span from=5, to=6, type=:punct>]
=end
It should be noted that it is also possible to process much complex texts than the present one.
text =<<-EOF
Hello! I am 18! My favourite number is 133.7...
What about you?
EOF
pp Greeb::Tokenizer.tokenize(text)
=begin
[#<struct Greeb::Span from=0, to=5, type=:letter>,
#<struct Greeb::Span from=5, to=6, type=:punct>,
#<struct Greeb::Span from=6, to=7, type=:space>,
#<struct Greeb::Span from=7, to=8, type=:letter>,
#<struct Greeb::Span from=8, to=9, type=:space>,
#<struct Greeb::Span from=9, to=11, type=:letter>,
#<struct Greeb::Span from=11, to=12, type=:space>,
#<struct Greeb::Span from=12, to=14, type=:integer>,
#<struct Greeb::Span from=14, to=15, type=:punct>,
#<struct Greeb::Span from=15, to=16, type=:space>,
#<struct Greeb::Span from=16, to=18, type=:letter>,
#<struct Greeb::Span from=18, to=19, type=:space>,
#<struct Greeb::Span from=19, to=28, type=:letter>,
#<struct Greeb::Span from=28, to=29, type=:space>,
#<struct Greeb::Span from=29, to=35, type=:letter>,
#<struct Greeb::Span from=35, to=36, type=:space>,
#<struct Greeb::Span from=36, to=38, type=:letter>,
#<struct Greeb::Span from=38, to=39, type=:space>,
#<struct Greeb::Span from=39, to=44, type=:float>,
#<struct Greeb::Span from=44, to=47, type=:punct>,
#<struct Greeb::Span from=47, to=49, type=:break>,
#<struct Greeb::Span from=49, to=53, type=:letter>,
#<struct Greeb::Span from=53, to=54, type=:space>,
#<struct Greeb::Span from=54, to=59, type=:letter>,
#<struct Greeb::Span from=59, to=60, type=:space>,
#<struct Greeb::Span from=60, to=63, type=:letter>,
#<struct Greeb::Span from=63, to=64, type=:punct>,
#<struct Greeb::Span from=64, to=65, type=:break>]
=end
The analyzer can also perform sentence detection.
text = 'Hello! How are you?'
tokens = Greeb::Tokenizer.tokenize(text)
pp Greeb::Segmentator.new(tokens).sentences
=begin
[#<struct Greeb::Span from=0, to=6, type=:sentence>,
#<struct Greeb::Span from=7, to=19, type=:sentence>]
=end
Having obtained the sentence boundaries, it is possible to extract tokens covered by these sentences.
text = 'Hello! How are you?'
tokens = Greeb::Tokenizer.tokenize(text)
segmentator = Greeb::Segmentator.new(tokens)
pp segmentator.extract(segmentator.sentences)
=begin
{#<struct Greeb::Span from=0, to=6, type=:sentence>=>
[#<struct Greeb::Span from=0, to=5, type=:letter>,
#<struct Greeb::Span from=5, to=6, type=:punct>],
#<struct Greeb::Span from=7, to=19, type=:sentence>=>
[#<struct Greeb::Span from=7, to=10, type=:letter>,
#<struct Greeb::Span from=10, to=11, type=:space>,
#<struct Greeb::Span from=11, to=14, type=:letter>,
#<struct Greeb::Span from=14, to=15, type=:space>,
#<struct Greeb::Span from=15, to=18, type=:letter>,
#<struct Greeb::Span from=18, to=19, type=:punct>]}
=end
It is often that a text includes such special entries as URLs and e-mail addresses. Greeb can assist you in extracting them.
text = 'My website is http://nlpub.ru and e-mail is example@example.com.'
pp Greeb::Parser.urls(text).map { |e| [e, e.slice(text)] }
=begin
[[#<struct Greeb::Span from=14, to=29, type=:url>, "http://nlpub.ru"]]
=end
pp Greeb::Parser.emails(text).map { |e| [e, e.slice(text)] }
=begin
[[#<struct Greeb::Span from=44, to=63, type=:email>, "example@example.com"]]
=end
Please do not use Greeb for the development of spam lists. Spam sucks.
text = 'Hello, G.L.H.F. everyone!'
pp Greeb::Parser.abbrevs(text).map { |e| [e, e.slice(text)] }
=begin
[[#<struct Greeb::Span from=7, to=15, type=:abbrev>, "G.L.H.F."]]
=end
The algorithm is not so accurate, but still useful in many practical situations.
text = 'Our time is running out: 13:37 or 14:89.'
pp Greeb::Parser.time(text).map { |e| [e, e.slice(text)] }
=begin
[[#<struct Greeb::Span from=25, to=30, type=:time>, "13:37"]]
=end
Greeb operates with spans, which are tuples of (from, to, kind), where from is a beginning of the span, to is an ending of the span, and kind is a type of the span.
There are several span types at the tokenization stage: :letter
,
:float
, :integer
, :separ
, :punct
(for punctuation), :spunct
(for in-sentence punctuation), :space
, and :break
.
- Fork it;
- Create your feature branch (
git checkout -b my-new-feature
); - Commit your changes (
git commit -am 'Added some feature'
); - Push to the branch (
git push origin my-new-feature
); - Create new Pull Request.
Copyright (c) 2010-2019 Dmitry Ustalov. See LICENSE for details.