wikipedia-vandalism_detection 0.0.1

data/lib/wikipedia/vandalism_detection/features/removed_markup_frequency.rb

@@ -0,0 +1,28 @@
+require 'wikipedia/vandalism_detection/features/base'
+require 'wikipedia/vandalism_detection/word_lists/markup'
+require 'wikipedia/vandalism_detection/text'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes the frequency of markup words in the removed text.
+      class RemovedMarkupFrequency < Base
+
+        # Returns the percentage of markup words in the removed text.
+        # Returns 0.0 if cleaned removed text is of zero length.
+        def calculate(edit)
+          super
+
+          text = edit.removed_text
+          all_words_count = edit.removed_words.count
+
+          regex = /(#{WordLists::MARKUP.join('|')})/
+          markup_words_count = text.scan(regex).count
+
+          (all_words_count > 0) ? (markup_words_count.to_f) / (all_words_count.to_f) : 0.0
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/removed_pronoun_frequency.rb

@@ -0,0 +1,23 @@
+require 'wikipedia/vandalism_detection/features/frequency_base'
+require 'wikipedia/vandalism_detection/word_lists/pronouns'
+require 'wikipedia/vandalism_detection/text'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes the frequency of pronouns in the removed text.
+      class RemovedPronounFrequency < FrequencyBase
+
+        # Returns the percentage of pronoun words in the removed text.
+        # Returns 0.0 if cleaned removed text is of zero length.
+        def calculate(edit)
+          super
+
+          text = Text.new(edit.removed_words.join("\n")).clean
+          frequency(text, WordLists::PRONOUNS)
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/removed_sex_frequency.rb

@@ -0,0 +1,23 @@
+require 'wikipedia/vandalism_detection/features/frequency_base'
+require 'wikipedia/vandalism_detection/word_lists/sex'
+require 'wikipedia/vandalism_detection/text'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes the frequency of sex words in the removed text.
+      class RemovedSexFrequency < FrequencyBase
+
+        # Returns the percentage of sex words in the removed text.
+        # Returns 0.0 if cleaned removed text is of zero length.
+        def calculate(edit)
+          super
+
+          text = Text.new(edit.removed_words.join("\n")).clean
+          frequency(text, WordLists::SEX)
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/removed_size.rb

@@ -0,0 +1,20 @@
+require 'wikipedia/vandalism_detection/features/base'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes the size of removed text in the edit's new revision.
+      class RemovedSize < Base
+
+        # Returns the size of removed character in the new revision.
+        def calculate(edit)
+          super
+
+          size = edit.removed_text.size
+          size >= 0 ? size : 0
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/removed_vulgarism_frequency.rb

@@ -0,0 +1,23 @@
+require 'wikipedia/vandalism_detection/features/frequency_base'
+require 'wikipedia/vandalism_detection/word_lists/vulgarism'
+require 'wikipedia/vandalism_detection/text'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes frequency of vulgarism words in the removed text.
+      class RemovedVulgarismFrequency < FrequencyBase
+
+        # Returns the percentage of vulgarism words in the removed text.
+        # Returns 0.0 if removed clean text is of zero length.
+        def calculate(edit)
+          super
+
+          text = Text.new(edit.removed_words.join("\n")).clean
+          frequency(text, WordLists::VULGARISM)
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/removed_words.rb

@@ -0,0 +1,18 @@
+require 'wikipedia/vandalism_detection/features/base'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes the number of removed words in the edit's new revision.
+      class RemovedWords < Base
+
+        def calculate(edit)
+          super
+
+          edit.removed_words.count
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/replacement_similarity.rb

@@ -0,0 +1,22 @@
+require 'wikipedia/vandalism_detection/features/base'
+require 'wikipedia/vandalism_detection/diff'
+require 'hotwater'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes the similarity of deleted to inserted text.
+      # As similarity measure the Jaro-Winkler distance is used.
+      # See: http://courses.cs.washington.edu/courses/cse590q/04au/papers/Winkler99.pdf
+      class ReplacementSimilarity < Base
+
+        def calculate(edit)
+          super
+
+          ::Hotwater.jaro_winkler_distance(edit.removed_text, edit.inserted_text)
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/reverted.rb

@@ -0,0 +1,18 @@
+require 'wikipedia/vandalism_detection/features/contains_base'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature returns whether the edit's comment includes 'copyedit'.
+      class Copyedit < ContainsBase
+
+        def calculate(edit)
+          super
+
+          contains(edit.new_revision.comment, ['copyedit', 'copy edit'])
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/revisions_character_distribution.rb

@@ -0,0 +1,21 @@
+require 'wikipedia/vandalism_detection/features/base'
+require 'wikipedia/vandalism_detection/algorithms'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes the Kullback-Leibler Divergence of the old and new text's character distribution.
+      # The smaller the divergence, the higher the similarity of the distributions and conversely.
+      class RevisionsCharacterDistribution < Base
+        include Algorithms
+
+        def calculate(edit)
+          super
+
+          kullback_leibler_divergence(edit.old_revision.text, edit.new_revision.text)
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/same_editor.rb

@@ -0,0 +1,29 @@
+require_relative 'base'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature describes whether the contributor of the new revision is the same as the editor
+      # of the old revision.
+      class SameEditor < Base
+
+        def calculate(edit)
+          super
+
+          old_revision = edit.old_revision
+
+          if old_revision.contributor.blank?
+            xml = Wikipedia::api_request({ prop: 'revisions', rvprop: 'user',revids: old_revision.id })
+            contributor = xml.xpath('//rev/@user').text
+            return Features::MISSING_VALUE if contributor.blank?
+
+            old_revision.contributor = contributor
+          end
+
+          (old_revision.contributor == edit.new_revision.contributor) ? 1 : 0
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/sex_frequency.rb

@@ -0,0 +1,23 @@
+require 'wikipedia/vandalism_detection/features/frequency_base'
+require 'wikipedia/vandalism_detection/word_lists/sex'
+require 'wikipedia/vandalism_detection/text'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes frequency of bad words in the inserted text.
+      class SexFrequency < FrequencyBase
+
+        # Returns the percentage of biased words in the inserted text.
+        # Returns 0.0 if inserted clean text is of zero length.
+        def calculate(edit)
+          super
+
+          text = Text.new(edit.inserted_words.join("\n")).clean
+          frequency(text, WordLists::SEX)
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/sex_impact.rb

@@ -0,0 +1,21 @@
+require 'wikipedia/vandalism_detection/features/impact_base'
+require 'wikipedia/vandalism_detection/word_lists/sex'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes the percentage by which the edit increases the number of sex words in the text.
+      class SexImpact < ImpactBase
+
+        def calculate(edit)
+          super
+          old_text = edit.old_revision.text.clean
+          new_text = edit.new_revision.text.clean
+
+          impact(old_text, new_text, WordLists::SEX)
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/size_increment.rb

@@ -0,0 +1,22 @@
+require 'wikipedia/vandalism_detection/features/base'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes the increment of the edit's revisions text length.
+      class SizeIncrement < Base
+
+        #	computation: |new| - |old|
+        def calculate(edit)
+          super
+
+          old_size = edit.old_revision.text.size
+          new_size = edit.new_revision.text.size
+
+          new_size - old_size
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/size_ratio.rb

@@ -0,0 +1,26 @@
+require 'wikipedia/vandalism_detection/features/base'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes the ratio of the edit's revisions text length.
+      class SizeRatio < Base
+
+        # Returns the ration of new text length to old text length:
+        # returns 0.0 for empty old revision text,
+        # returns 1.0 for empty new revision text,
+        # returns 0.5 for both revision texts empty or same size
+        #	computation: old / old + new
+        def calculate(edit)
+          super
+
+          old_size = edit.old_revision.text.size.to_f
+          new_size = edit.new_revision.text.size.to_f
+
+          (old_size == 0 && new_size == 0) ? 0.5 : old_size / (old_size + new_size)
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/term_frequency.rb

@@ -0,0 +1,25 @@
+require 'wikipedia/vandalism_detection/features/base'
+require 'wikipedia/vandalism_detection/text'
+require 'wikipedia/vandalism_detection/diff'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes average frequency of words inserted in the new revision
+      # relative to the words in the old revision.
+      class TermFrequency < FrequencyBase
+
+        def calculate(edit)
+          super
+
+          new_text = edit.new_revision.text
+          inserted_terms = Text.new(edit.inserted_words.join("\n")).clean.gsub(/[^\w\s]/, '').split.uniq
+          summed_frequencies = inserted_terms.reduce(0) { |count, term| count + frequency(new_text.clean, term) }
+
+          (inserted_terms.count > 0) ? (summed_frequencies / inserted_terms.count) : 0.0
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/time_interval.rb

@@ -0,0 +1,31 @@
+require 'wikipedia/vandalism_detection/features/base'
+require 'date'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes the time interval in days between old and new revision.
+      class TimeInterval < Base
+
+        def calculate(edit)
+          super
+
+          new_time = DateTime.parse(edit.new_revision.timestamp)
+
+          if edit.old_revision.timestamp.blank?
+            xml = Wikipedia::api_request({ prop: 'revisions', rvprop: 'timestamp', revids: edit.old_revision.id })
+            timestamp = xml.xpath('//rev/@timestamp').text
+            return Features::MISSING_VALUE if timestamp.blank?
+
+            old_time = DateTime.parse(timestamp)
+          else
+            old_time = DateTime.parse(edit.old_revision.timestamp)
+          end
+
+          (new_time - old_time).to_f.abs
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/time_of_day.rb

@@ -0,0 +1,22 @@
+require 'wikipedia/vandalism_detection/features/base'
+require 'date'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature calculates the time of day of new revision edit as decimal value .
+      class TimeOfDay < Base
+
+        def calculate(edit)
+          super
+
+          timestamp = edit.new_revision.timestamp
+          time = DateTime.parse(timestamp)
+
+          time.hour.to_f + (time.min / 60.0) + (time.sec / 360.0)
+        end
+      end
+    end
+  end
+end

data/lib/wikipedia/vandalism_detection/features/upper_case_ratio.rb

@@ -0,0 +1,24 @@
+require 'wikipedia/vandalism_detection/features/base'
+
+module Wikipedia
+  module VandalismDetection
+    module Features
+
+      # This feature computes the uppercase to all letters ratio of the edit's new revision inserted text.
+      class UpperCaseRatio < Base
+
+        def calculate(edit)
+          super
+
+          text = edit.inserted_text
+          return 0.0 if text.empty?
+
+          uppercase_count = text.scan(/[[:upper:]]/).size
+          all_letters_count = text.scan(/[[:alpha:]]/).size
+
+          (1.0 + uppercase_count) / (1.0 + all_letters_count)
+        end
+      end
+    end
+  end
+end