Fix negator token lookup on sentiment analysis

spec/cadmium/i18n/stop_words_spec.cr

@@ -2,15 +2,11 @@ include Cadmium::I18n::StopWords
 stop_words en, fr
 stop_words all_languages
 describe Cadmium::I18n::StopWords do
-  subject = stop_words_fr
-
   it "should return a word from the french stop words list" do
-    subject[2].should eq("absolument")
+    stop_words_fr[2].should eq("absolument")
   end
 
-  subject = stop_words_all_languages["ru"]
-
   it "should return a word from the russian stop words list" do
-    subject[45].should eq("взгляд")
+    stop_words_all_languages["ru"][45].should eq("взгляд")
   end
 end

src/cadmium/readability.cr

@@ -74,7 +74,7 @@ module Cadmium
     # The average number of sentences per 100 words. Useful for the Coleman-Liau
     # and Linsear Write score calculation
     def sentences_per_hundred_words
-      sentences.size.to_f / (words.size / 100).to_f
+      sentences.size.to_f / (words.size // 100).to_f
     end
 
     # The average number of characters per word. Useful for the Coleman-Liau
@@ -133,7 +133,7 @@ module Cadmium
     # ARI uses a scale based on age in full-time education.
 
     def ari
-      result = 4.71 * (num_chars / num_words) + 0.5 * (num_words / num_sentences) - 21.43
+      result = 4.71 * (num_chars // num_words) + 0.5 * (num_words // num_sentences) - 21.43
       result.finite? ? result.round(2) : 0.0
     end
 
@@ -156,7 +156,7 @@ module Cadmium
     # good standard for ordinary text.
 
     def lix
-      result = (num_words / num_sentences).to_f + ((@long_words * 100) / num_words).to_f
+      result = (num_words // num_sentences).to_f + ((@long_words * 100) // num_words).to_f
       result.finite? ? result.round(2) : 0.0
     end
 

src/cadmium/sentiment.cr

@@ -6,20 +6,20 @@ module Cadmium
 
     # Negate the next word in the phrase.
     NEGATORS = {
-      "cant"    => 1,
-      "can't"   => 1,
-      "dont"    => 1,
-      "don't"   => 1,
-      "doesnt"  => 1,
-      "doesn't" => 1,
-      "not"     => 1,
-      "non"     => 1,
-      "wont"    => 1,
-      "won't"   => 1,
-      "isnt"    => 1,
-      "isn't"   => 1,
-      "wasnt"   => 1,
-      "wasn't"  => 1,
+      "cant",
+      "can't",
+      "dont",
+      "don't",
+      "doesnt",
+      "doesn't",
+      "not",
+      "non",
+      "wont",
+      "won't",
+      "isnt",
+      "isn't",
+      "wasnt",
+      "wasn't",
     }
 
     # Manage the `Tokenizer` that the sentiment analyzer uses.

src/cadmium/summarizer/luhn_summarizer.cr

@@ -33,10 +33,10 @@ module Cadmium
       window_size = window_size(terms_in_sentence, normalized_terms)
       return 0 if window_size <= 0
       number_of_normalized_terms = terms_in_sentence.count { |term| normalized_terms.includes?(term) }
-      (number_of_normalized_terms*number_of_normalized_terms) / window_size
+      (number_of_normalized_terms*number_of_normalized_terms) // window_size
     end
 
-    private def select_sentences(text, max_num_sentences, normalized_terms_ratio)
+    private def select_sentences(text, max_num_sentences, normalized_terms_ratio) : Array(String)
       sentences = Cadmium::Util::Sentence.sentences(text)
       sentences.sort_by! { |sentence| -sentence_rating(sentence, normalized_terms_ratio) } # This could be improved, performance wise.
       sentences[0..max_num_sentences]