TactfulTokenizer is a Ruby library for high quality sentence tokenization. It uses a Naive Bayesian statistical model, and is based on Splitta, but has support for '?' and '!' as well as primitive handling of XHTML markup. Better support for XHTML parsing is coming shortly.
Additionally supports unicode text tokenization.
require "tactful_tokenizer" m = TactfulTokenizer::Model.new m.tokenize_text("Here in the U.S. Senate we prefer to eat our friends. Is it easier that way? <em>Yes.</em> <em>Maybe</em>!") #=> ["Here in the U.S. Senate we prefer to eat our friends.", "Is it easier that way?", "<em>Yes.</em>", "<em>Maybe</em>!"]
The input text is expected to consist of paragraphs delimited by line breaks.
gem install tactful_tokenizer
Copyright © 2010 Matthew Bunday. All rights reserved. Released under the GNU GPL v3.