Use UNICODE-compatible regular expressions.

Even though english was enough for the lord to write the bible, it's still a smart idea to allow for UNICODE characters if only to allow horses to coöperate in a way the New Yorker would approve.

See:
http://www.ruby-doc.org/core-1.9.3/Regexp.html#label-Character+Properties 
and 
http://www.newyorker.com/online/blogs/culture/2012/04/the-curse-of-the-diaeresis.html

ebook.rb

@@ -60,7 +60,7 @@ else
   source_tweets.each do |twt|
     text = twt
 
-    if text !~ /[\.\"\'\?\!]/
+    if text !~ /\p{Punct}/
       text += "."
     end
     
@@ -72,9 +72,9 @@ else
   10.times do
     tweet = markov.generate_sentence
 
-    if rand(3) == 0 && tweet =~ /(in|to|from|for|with|by|our|of|your|around|under|beyond)\s\w+$/ 
+    if rand(3) == 0 && tweet =~ /(in|to|from|for|with|by|our|of|your|around|under|beyond)\p{Space}\w+$/ 
       puts "Losing last word randomly"
-      tweet.gsub(/\s\w+.$/, '')   # randomly losing the last word sometimes like horse_ebooks
+      tweet.gsub(/\p{Space}\p{Word}+.$/, '')   # randomly losing the last word sometimes like horse_ebooks
     end
 
     if tweet.length < 40 && rand(5) == 0
@@ -86,8 +86,8 @@ else
       puts "MARKOV: #{tweet}"
     end
 
-    tweet_letters = tweet.gsub(/\W/, '')
-    break if !tweet.nil? && tweet.length < 110 && !source_tweets.any? {|t| t.gsub(/\W/, '') =~ /#{tweet_letters}/ }
+    tweet_letters = tweet.gsub(/\P{Word}/, '')
+    break if !tweet.nil? && tweet.length < 110 && !source_tweets.any? {|t| t.gsub(/\P{Word}/, '') =~ /#{tweet_letters}/ }
   end
   
   if params["tweet"]

markov.rb

@@ -10,9 +10,9 @@ class MarkovChainer
 
    def add_text(text)
      # make sure each paragraph ends with some sentence terminator
-     text.gsub!(/\n\s*\n/m, ".")
+     text.gsub!(/[\r\n]+\p{Space}*[\r\n]+/m, ".")
      text << "."
-     seps = /([.!?;])/
+     seps = /(\p{Punct})/
      sentence = ""
      text.split(seps).each { |p|
        if seps =~ p
@@ -37,7 +37,7 @@ class MarkovChainer
 
 private
    def add_sentence(str, terminator)
-     words = str.scan(/[\w'’\-]+/)
+     words = str.scan(/[\p{Word}'’\-]+/)
      return unless words.size > order # ignore short sentences
      words << terminator
      buf = []