wikipedia-vandalism_detection 0.0.1

data/spec/vandalism_detection/features/vulgarism_frequency_spec.rb

@@ -0,0 +1,36 @@
+require 'spec_helper'
+
+describe Wikipedia::VandalismDetection::Features::VulgarismFrequency do
+
+  before do
+    @feature = Wikipedia::VandalismDetection::Features::VulgarismFrequency.new
+  end
+
+  it { should be_a Wikipedia::VandalismDetection::Features::FrequencyBase }
+
+  describe "#calculate" do
+
+    it "returns the number of vulgarism words relative to all words count" do
+      # total 10 words, 3 vulgarism
+      old_text = Wikipedia::VandalismDetection::Text.new 'Old shit.'
+      new_text = Wikipedia::VandalismDetection::Text.new "Old shit. Fuck you, fu*ck you $lut, you and all the others."
+
+      old_revision = build(:old_revision, text: old_text)
+      new_revision = build(:new_revision, text: new_text)
+      edit = build(:edit, old_revision: old_revision, new_revision: new_revision)
+
+      expect(@feature.calculate(edit)).to eq 3.0/10.0
+    end
+
+    it "returns 0.0 on emtpy clean text revisions" do
+      old_text = Wikipedia::VandalismDetection::Text.new 'Old shit.'
+      new_text = Wikipedia::VandalismDetection::Text.new "Old shit. {{speedy deletion}}"
+
+      old_revision = build(:old_revision, text: old_text)
+      new_revision = build(:new_revision, text: new_text)
+      edit = build(:edit, old_revision: old_revision, new_revision: new_revision)
+
+      expect(@feature.calculate(edit)).to eq 0.0
+    end
+  end
+end

data/spec/vandalism_detection/features/vulgarism_impact_spec.rb

@@ -0,0 +1,58 @@
+require 'spec_helper'
+
+describe  Wikipedia::VandalismDetection::Features::VulgarismImpact do
+
+  before do
+    @feature = Wikipedia::VandalismDetection::Features::VulgarismImpact.new
+  end
+
+  it { should be_a Wikipedia::VandalismDetection::Features::ImpactBase }
+
+  describe "#calculate" do
+
+    it "returns the impact of vulgarism words of the edit's new revision text" do
+      # 1 vulgarism
+      old_text = Wikipedia::VandalismDetection::Text.new "Fuck you, you and all the others"
+      # 3 vulgarism
+      new_text = Wikipedia::VandalismDetection::Text.new "Fuck you, fuck you, you and all the others sluts"
+
+      old_revision = build(:old_revision, text: old_text)
+      new_revision = build(:new_revision, text: new_text)
+      edit = build(:edit, old_revision: old_revision, new_revision: new_revision)
+
+      expect(@feature.calculate(edit)).to eq 1.0 / (1.0 + 3.0)
+    end
+
+    it "returns 0.5 on both no terms in text revisions" do
+      text = Wikipedia::VandalismDetection::Text.new "{{speedy deletion}}"
+
+      old_revision = build(:old_revision, text: text)
+      new_revision = build(:new_revision, text: text)
+      edit = build(:edit, new_revision: new_revision, old_revision: old_revision)
+
+      expect(@feature.calculate(edit)).to eq 0.5
+    end
+
+    it "returns 0.0 on emtpy clean text of old revision" do
+      old_text = Wikipedia::VandalismDetection::Text.new "{{speedy deletion}}"
+      new_text = Wikipedia::VandalismDetection::Text.new "fuck"
+
+      old_revision = build(:old_revision, text: old_text)
+      new_revision = build(:new_revision, text: new_text)
+      edit = build(:edit, new_revision: new_revision, old_revision: old_revision)
+
+      expect(@feature.calculate(edit)).to eq 0.0
+    end
+
+    it "returns 1.0 on emtpy clean text of new revision" do
+      old_text = Wikipedia::VandalismDetection::Text.new "fuck"
+      new_text = Wikipedia::VandalismDetection::Text.new "{{speedy deletion}}"
+
+      old_revision = build(:old_revision, text: old_text)
+      new_revision = build(:new_revision, text: new_text)
+      edit = build(:edit, new_revision: new_revision, old_revision: old_revision)
+
+      expect(@feature.calculate(edit)).to eq 1.0
+    end
+  end
+end

data/spec/vandalism_detection/features/weekday_spec.rb

@@ -0,0 +1,22 @@
+require 'spec_helper'
+
+describe Wikipedia::VandalismDetection::Features::Weekday do
+
+  before do
+    @feature = Wikipedia::VandalismDetection::Features::Weekday.new
+  end
+
+  it { should be_a Wikipedia::VandalismDetection::Features::Base }
+
+  describe "#calculate" do
+
+    it "returns the weekday as decimal value" do
+      old_revision = build(:old_revision)
+      new_revision = build(:new_revision, timestamp: '2012-12-11T05:30:36Z' )
+
+      edit = build(:edit, old_revision: old_revision, new_revision: new_revision)
+
+      expect(@feature.calculate(edit)).to eq 2 # Thuesday
+    end
+  end
+end

data/spec/vandalism_detection/features/words_increment_spec.rb

@@ -0,0 +1,35 @@
+require 'spec_helper'
+
+describe Wikipedia::VandalismDetection::Features::WordsIncrement do
+
+  before do
+    @feature = Wikipedia::VandalismDetection::Features::WordsIncrement.new
+  end
+
+  it { should be_a Wikipedia::VandalismDetection::Features::Base }
+
+  describe "#calculate" do
+
+    it "returns a negative increment on more removed texts" do
+      old_revision_text = Wikipedia::VandalismDetection::Text.new 'one two three four five six seven eight nine' # length 9
+      new_revision_text = Wikipedia::VandalismDetection::Text.new 'one two three' # length 3
+
+      old_revision = build(:old_revision, text: old_revision_text)
+      new_revision = build(:new_revision, text: new_revision_text)
+      edit = build(:edit, old_revision: old_revision, new_revision: new_revision)
+
+      expect(@feature.calculate(edit)).to eq 3.0 - 9.0
+    end
+
+    it "returns a positive increment on more removed texts" do
+      old_revision_text = Wikipedia::VandalismDetection::Text.new 'one two three' # length 3
+      new_revision_text = Wikipedia::VandalismDetection::Text.new 'one two three four five six seven eight nine' # length 9
+
+      old_revision = build(:old_revision, text: old_revision_text)
+      new_revision = build(:new_revision, text: new_revision_text)
+      edit = build(:edit, old_revision: old_revision, new_revision: new_revision)
+
+      expect(@feature.calculate(edit)).to eq 9.0 - 3.0
+    end
+  end
+end

data/spec/vandalism_detection/instances_spec.rb

@@ -0,0 +1,156 @@
+require 'spec_helper'
+require 'wikipedia/vandalism_detection/instances'
+
+describe Wikipedia::VandalismDetection::Instances do
+
+  it "responds to #empty" do
+    expect { Wikipedia::VandalismDetection::Instances.empty }.not_to raise_error
+  end
+
+  describe "#empty" do
+
+    before do
+      use_test_configuration
+
+      @dataset = Wikipedia::VandalismDetection::Instances.empty
+      @attributes = @dataset.enumerate_attributes
+      @class_attribute = @dataset.class_attribute
+    end
+
+    it "returns a weka dataset" do
+      expect(@dataset.class).to eq Java::WekaCore::Instances::Base
+    end
+
+    it "returns an empty dataset" do
+      expect(@dataset.n_rows).to eq 0
+    end
+
+    it "has all configured features and class as attributes" do
+      attribute_names = @attributes.map{ |attr| "#{attr.name.gsub('_', ' ')}" }
+      features = test_config.features
+
+      expect(attribute_names).to eq features
+    end
+
+    it "has feature attributes of type 'numeric'" do
+      feature_attributes = (@attributes.to_a)[0...-1]
+      all_features_numeric = feature_attributes.reduce{ |result, attr| result && attr.numeric? }
+
+      expect(all_features_numeric).to be true
+    end
+
+    it "has a class attribute of type 'nominal'" do
+      expect(@class_attribute.nominal?).to be true
+    end
+
+    it "has a class attribute with values 'vandalism' and 'regular'" do
+      values = @class_attribute.num_values.times.collect {|index| @class_attribute.value(index) }
+      expect(values).to eq ['regular','vandalism']
+    end
+  end
+
+  describe "#empty_for_feature" do
+    before do
+      @dataset = Wikipedia::VandalismDetection::Instances.empty_for_feature('comment length')
+      @attributes = @dataset.enumerate_attributes
+      @class_attribute = @dataset.class_attribute
+    end
+
+    it "returns a weka dataset" do
+      expect(@dataset.class).to be Java::WekaCore::Instances::Base
+    end
+
+    it "returns an empty dataset" do
+      expect(@dataset.n_rows).to eq 0
+    end
+
+    it "has only given feature and class as attributes" do
+      attribute_names = @attributes.map{ |attr| "#{attr.name.gsub('_', ' ')}" }
+      features = ['comment length']
+
+      expect(attribute_names).to eq features
+    end
+
+    it "has feature attributes of type 'numeric'" do
+      attribute = (@attributes.to_a)[0]
+      expect(attribute.numeric?).to be true
+    end
+
+    it "has a class attribute of type 'nominal'" do
+      expect(@class_attribute.nominal?).to be true
+    end
+
+    it "has a class attribute with values 'vandalism' and 'regular'" do
+      values = @class_attribute.num_values.times.collect {|index| @class_attribute.value(index) }
+      expect(values).to eq ['regular','vandalism']
+    end
+  end
+
+  describe "#empty_for_test_feature" do
+
+    before do
+      @dataset = Wikipedia::VandalismDetection::Instances.empty_for_test_feature('comment length')
+
+      @feature_attribute = @dataset.enumerate_attributes.to_a.first
+      @old_revision_id_attribute = @dataset.enumerate_attributes.to_a[-2]
+      @new_revision_id_attribute = @dataset.enumerate_attributes.to_a.last
+    end
+
+    it "returns a weka dataset" do
+      expect(@dataset.class).to be Java::WekaCore::Instances::Base
+    end
+
+    it "returns an empty dataset" do
+      expect(@dataset.n_rows).to eq 0
+    end
+
+    it "has one given feature as attributes" do
+      attribute_name = @feature_attribute.name.gsub('_', ' ')
+      expect(attribute_name).to eq 'comment length'
+    end
+
+    it "has feature attributes of type 'numeric'" do
+      expect(@feature_attribute.numeric?).to be true
+    end
+
+    it "has an attribute with name 'oldrevisionid'" do
+      expect(@old_revision_id_attribute.name).to  eq 'oldrevisionid'
+    end
+
+    it "has an oldrevisionid attribute of type 'numeric'" do
+      expect(@old_revision_id_attribute.numeric?).to be true
+    end
+
+    it "has an attribute with name 'newrevisionid'" do
+      expect(@new_revision_id_attribute.name).to eq 'newrevisionid'
+    end
+
+    it "has a newrevisionid attribute of type 'numeric'" do
+      expect(@new_revision_id_attribute.numeric?).to be true
+    end
+  end
+
+  describe "#empty_for_test_class" do
+
+    before do
+      @dataset = Wikipedia::VandalismDetection::Instances.empty_for_test_class
+      @class = @dataset.enumerate_attributes.to_a.first
+    end
+
+    it "returns a weka dataset" do
+      expect(@dataset.class).to be Java::WekaCore::Instances::Base
+    end
+
+    it "returns an empty dataset" do
+      expect(@dataset.n_rows).to eq 0
+    end
+
+    it "has one given feature as attributes" do
+      expect(@class.name).to eq 'class'
+    end
+
+    it "has feature attributes of type 'nominal'" do
+      expect(@class.nominal?).to be true
+    end
+  end
+end

data/spec/vandalism_detection/page_parser_spec.rb

@@ -0,0 +1,184 @@
+require 'spec_helper'
+
+describe Wikipedia::VandalismDetection::PageParser do
+
+  before do
+    @parser =  Wikipedia::VandalismDetection::PageParser.new
+  end
+
+  describe "parser structure" do
+
+    describe "#parse" do
+      before do
+        xml = load_file('vandalism_on_wikipedia.xml')
+        @page = @parser.parse(xml)
+      end
+
+      it "returns a Wikipedia::Page object" do
+        expect(@page).to be_a Wikipedia::VandalismDetection::Page
+      end
+
+      it "returns a Wikipedia::Page with the right number of revisions" do
+        expect(@page.revisions.count).to eq 5
+      end
+    end
+  end
+
+  describe "a single page content parsing" do
+    before do
+      @xml = load_file('vandalism_on_wikipedia.xml')
+      @page = @parser.parse @xml
+
+      revision_1 = build :empty_revision, id: '398880281'
+      revision_2 = build :empty_revision, id: '398880502', parent_id: '398880281'
+      revision_3 = build :empty_revision, id: '398883278', parent_id: '398880502'
+      revision_4 = build :empty_revision, id: '398883675', parent_id: '398883278'
+      revision_5 = build :empty_revision, id: '398885233', parent_id: '398883675'
+
+      @revisions = {revision_1.id => revision_1, revision_2.id => revision_2, revision_3.id => revision_3,
+                    revision_4.id => revision_4, revision_5.id => revision_5}
+    end
+
+    it "can read a single page dump text input" do
+      expect(@page).to_not be_nil
+    end
+
+    it 'has a title' do
+      expect(@page.title).to eq 'Vandalism on Wikipedia'
+    end
+
+    it 'has an id' do
+      expect(@page.id).to eq '29753790'
+    end
+
+
+    describe "page's revisions" do
+      it 'has the right number of revisions' do
+        expect(@page.revisions.count).to eq 5
+      end
+
+      it "has revisions each with the right id" do
+        page_revisions = @page.revisions
+
+        page_revisions.each do |key, value|
+          expect(value.id).to eq @revisions[key].id
+        end
+      end
+
+      it "has revisions each with the right parent_id" do
+        page_revisions = @page.revisions
+
+        page_revisions.each do |key, value|
+          expect(value.parent_id).to eq @revisions[key].parent_id
+        end
+      end
+
+      it "has revisions each with the right text" do
+        xml = load_file('vandalism_on_wikipedia_simplified.xml')
+        page = @parser.parse xml
+
+        revision_1 = build :empty_revision, id: '1', text: "text\n\n\n        1\n\n      "
+        revision_2 = build :empty_revision, id: '2', text: 'text 2'
+        revision_3 = build :empty_revision, id: '3', text: 'text 3'
+        revision_4 = build :empty_revision, id: '4', text: 'text 4'
+        revision_5 = build :empty_revision, id: '5', text: 'text 5'
+
+        revisions = {revision_1.id => revision_1, revision_2.id => revision_2, revision_3.id => revision_3,
+                      revision_4.id => revision_4, revision_5.id => revision_5}
+
+        page.revisions.each do |key, value|
+          expect(value.text).to eq revisions[key].text
+        end
+      end
+
+      it "has revisions each with the right comment" do
+        xml = load_file('vandalism_on_wikipedia_simplified.xml')
+        page = @parser.parse xml
+
+        revision_1 = build :empty_revision, id: '1', comment: "comment\n\n        1\n\n      "
+        revision_2 = build :empty_revision, id: '2', comment: 'comment 2'
+        revision_3 = build :empty_revision, id: '3'
+        revision_4 = build :empty_revision, id: '4'
+        revision_5 = build :empty_revision, id: '5', comment: 'comment 3'
+
+        revisions = {revision_1.id => revision_1, revision_2.id => revision_2, revision_3.id => revision_3,
+                     revision_4.id => revision_4, revision_5.id => revision_5}
+
+        page.revisions.each do |key, value|
+          expect(value.comment).to eq revisions[key].comment
+        end
+      end
+
+      describe "contributor properties" do
+        before do
+          xml = load_file('vandalism_on_wikipedia_simplified.xml')
+          @page = @parser.parse xml
+
+          revision_1 = build :empty_revision, id: '1', contributor: '1'
+          revision_2 = build :empty_revision, id: '2', contributor: '10', contributor_username: 'user'
+          revision_3 = build :empty_revision, id: '3', contributor: '11', contributor_username: 'user'
+          revision_4 = build :empty_revision, id: '4', contributor: '12', contributor_username: 'user'
+          revision_5 = build :empty_revision, id: '5', contributor: '2'
+
+          @revisions = {revision_1.id => revision_1, revision_2.id => revision_2, revision_3.id => revision_3,
+                       revision_4.id => revision_4, revision_5.id => revision_5}
+        end
+
+        it "has revisions each with the right contributor id" do
+          @page.revisions.each do |key, value|
+            expect(value.contributor_id).to eq @revisions[key].contributor_id
+          end
+        end
+
+        it "has revisions each with the right contributor ip" do
+          @page.revisions.each do |key, value|
+            expect(value.contributor_ip).to eq @revisions[key].contributor_ip
+          end
+        end
+
+        it "has revisions each with the right contributor username" do
+          @page.revisions.each do |key, value|
+            expect(value.contributor_username).to eq @revisions[key].contributor_username
+          end
+        end
+      end
+
+      it "has revisions each with the right timestamp" do
+        xml = load_file('vandalism_on_wikipedia_simplified.xml')
+        page = @parser.parse xml
+
+        revision_1 = build :empty_revision, id: '1', timestamp: 'time 1'
+        revision_2 = build :empty_revision, id: '2', timestamp: 'time 2'
+        revision_3 = build :empty_revision, id: '3', timestamp: 'time 3'
+        revision_4 = build :empty_revision, id: '4', timestamp: 'time 4'
+        revision_5 = build :empty_revision, id: '5', timestamp: 'time 5'
+
+        revisions = {revision_1.id => revision_1, revision_2.id => revision_2, revision_3.id => revision_3,
+                     revision_4.id => revision_4, revision_5.id => revision_5}
+
+        page.revisions.each do |key, value|
+          expect(value.timestamp).to eq revisions[key].timestamp
+        end
+      end
+
+      it "has revisions each with the right sha1 hash" do
+        xml = load_file('vandalism_on_wikipedia_simplified.xml')
+        page = @parser.parse xml
+
+        revision_1 = build :empty_revision, id: '1', sha1: 'hash1'
+        revision_2 = build :empty_revision, id: '2', sha1: 'hash2'
+        revision_3 = build :empty_revision, id: '3', sha1: 'hash3'
+        revision_4 = build :empty_revision, id: '4', sha1: 'hash4'
+        revision_5 = build :empty_revision, id: '5', sha1: 'hash5'
+
+        revisions = {revision_1.id => revision_1, revision_2.id => revision_2, revision_3.id => revision_3,
+                     revision_4.id => revision_4, revision_5.id => revision_5}
+
+        page.revisions.each do |key, value|
+          expect(value.sha1).to eq revisions[key].sha1
+        end
+      end
+
+    end
+  end
+end