wikipedia-vandalism_detection 0.0.1
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +7 -0
- data/.gitignore +17 -0
- data/Gemfile +6 -0
- data/LICENSE.txt +4 -0
- data/README.md +265 -0
- data/Rakefile +12 -0
- data/lib/java/LibSVM.jar +0 -0
- data/lib/java/SMOTE.jar +0 -0
- data/lib/java/balancedRandomForest.jar +0 -0
- data/lib/java/diffutils-1.3.0.jar +0 -0
- data/lib/java/libsvm.jar +0 -0
- data/lib/java/oneClassClassifier.jar +0 -0
- data/lib/java/realAdaBoost.jar +0 -0
- data/lib/java/swc-engine-1.1.0-jar-with-dependencies.jar +0 -0
- data/lib/java/sweble-wikitext-extractor.jar +0 -0
- data/lib/weka/classifiers/functions/lib_svm.rb +15 -0
- data/lib/weka/classifiers/meta/one_class_classifier.rb +25 -0
- data/lib/weka/classifiers/meta/real_ada_boost.rb +17 -0
- data/lib/weka/classifiers/trees/balanced_random_forest.rb +18 -0
- data/lib/weka/filters/supervised/instance/smote.rb +22 -0
- data/lib/wikipedia.rb +51 -0
- data/lib/wikipedia/vandalism_detection.rb +30 -0
- data/lib/wikipedia/vandalism_detection/algorithms.rb +18 -0
- data/lib/wikipedia/vandalism_detection/algorithms/kullback_leibler_divergence.rb +69 -0
- data/lib/wikipedia/vandalism_detection/classifier.rb +186 -0
- data/lib/wikipedia/vandalism_detection/configuration.rb +321 -0
- data/lib/wikipedia/vandalism_detection/diff.rb +27 -0
- data/lib/wikipedia/vandalism_detection/edit.rb +75 -0
- data/lib/wikipedia/vandalism_detection/evaluator.rb +606 -0
- data/lib/wikipedia/vandalism_detection/exceptions.rb +40 -0
- data/lib/wikipedia/vandalism_detection/feature_calculator.rb +89 -0
- data/lib/wikipedia/vandalism_detection/features.rb +67 -0
- data/lib/wikipedia/vandalism_detection/features/all_wordlists_frequency.rb +23 -0
- data/lib/wikipedia/vandalism_detection/features/all_wordlists_impact.rb +22 -0
- data/lib/wikipedia/vandalism_detection/features/anonymity.rb +19 -0
- data/lib/wikipedia/vandalism_detection/features/anonymity_previous.rb +29 -0
- data/lib/wikipedia/vandalism_detection/features/article_size.rb +18 -0
- data/lib/wikipedia/vandalism_detection/features/bad_frequency.rb +23 -0
- data/lib/wikipedia/vandalism_detection/features/bad_impact.rb +21 -0
- data/lib/wikipedia/vandalism_detection/features/base.rb +54 -0
- data/lib/wikipedia/vandalism_detection/features/biased_frequency.rb +23 -0
- data/lib/wikipedia/vandalism_detection/features/biased_impact.rb +22 -0
- data/lib/wikipedia/vandalism_detection/features/blanking.rb +25 -0
- data/lib/wikipedia/vandalism_detection/features/character_diversity.rb +25 -0
- data/lib/wikipedia/vandalism_detection/features/character_sequence.rb +19 -0
- data/lib/wikipedia/vandalism_detection/features/comment_bad_frequency.rb +22 -0
- data/lib/wikipedia/vandalism_detection/features/comment_biased_frequency.rb +22 -0
- data/lib/wikipedia/vandalism_detection/features/comment_length.rb +17 -0
- data/lib/wikipedia/vandalism_detection/features/comment_markup_frequency.rb +27 -0
- data/lib/wikipedia/vandalism_detection/features/comment_pronoun_frequency.rb +22 -0
- data/lib/wikipedia/vandalism_detection/features/comment_sex_frequency.rb +22 -0
- data/lib/wikipedia/vandalism_detection/features/comment_vulgarism_frequency.rb +22 -0
- data/lib/wikipedia/vandalism_detection/features/compressibility.rb +27 -0
- data/lib/wikipedia/vandalism_detection/features/contains_base.rb +18 -0
- data/lib/wikipedia/vandalism_detection/features/copyedit.rb +18 -0
- data/lib/wikipedia/vandalism_detection/features/digit_ratio.rb +24 -0
- data/lib/wikipedia/vandalism_detection/features/edits_per_user.rb +65 -0
- data/lib/wikipedia/vandalism_detection/features/emoticons_frequency.rb +27 -0
- data/lib/wikipedia/vandalism_detection/features/emoticons_impact.rb +29 -0
- data/lib/wikipedia/vandalism_detection/features/frequency_base.rb +20 -0
- data/lib/wikipedia/vandalism_detection/features/impact_base.rb +22 -0
- data/lib/wikipedia/vandalism_detection/features/inserted_character_distribution.rb +22 -0
- data/lib/wikipedia/vandalism_detection/features/inserted_external_links.rb +18 -0
- data/lib/wikipedia/vandalism_detection/features/inserted_internal_links.rb +18 -0
- data/lib/wikipedia/vandalism_detection/features/inserted_size.rb +20 -0
- data/lib/wikipedia/vandalism_detection/features/inserted_words.rb +18 -0
- data/lib/wikipedia/vandalism_detection/features/longest_word.rb +20 -0
- data/lib/wikipedia/vandalism_detection/features/markup_frequency.rb +27 -0
- data/lib/wikipedia/vandalism_detection/features/markup_impact.rb +27 -0
- data/lib/wikipedia/vandalism_detection/features/non_alphanumeric_ratio.rb +24 -0
- data/lib/wikipedia/vandalism_detection/features/personal_life.rb +18 -0
- data/lib/wikipedia/vandalism_detection/features/pronoun_frequency.rb +23 -0
- data/lib/wikipedia/vandalism_detection/features/pronoun_impact.rb +21 -0
- data/lib/wikipedia/vandalism_detection/features/removed_all_wordlists_frequency.rb +23 -0
- data/lib/wikipedia/vandalism_detection/features/removed_bad_frequency.rb +23 -0
- data/lib/wikipedia/vandalism_detection/features/removed_biased_frequency.rb +23 -0
- data/lib/wikipedia/vandalism_detection/features/removed_character_distribution.rb +22 -0
- data/lib/wikipedia/vandalism_detection/features/removed_emoticons_frequency.rb +27 -0
- data/lib/wikipedia/vandalism_detection/features/removed_markup_frequency.rb +28 -0
- data/lib/wikipedia/vandalism_detection/features/removed_pronoun_frequency.rb +23 -0
- data/lib/wikipedia/vandalism_detection/features/removed_sex_frequency.rb +23 -0
- data/lib/wikipedia/vandalism_detection/features/removed_size.rb +20 -0
- data/lib/wikipedia/vandalism_detection/features/removed_vulgarism_frequency.rb +23 -0
- data/lib/wikipedia/vandalism_detection/features/removed_words.rb +18 -0
- data/lib/wikipedia/vandalism_detection/features/replacement_similarity.rb +22 -0
- data/lib/wikipedia/vandalism_detection/features/reverted.rb +18 -0
- data/lib/wikipedia/vandalism_detection/features/revisions_character_distribution.rb +21 -0
- data/lib/wikipedia/vandalism_detection/features/same_editor.rb +29 -0
- data/lib/wikipedia/vandalism_detection/features/sex_frequency.rb +23 -0
- data/lib/wikipedia/vandalism_detection/features/sex_impact.rb +21 -0
- data/lib/wikipedia/vandalism_detection/features/size_increment.rb +22 -0
- data/lib/wikipedia/vandalism_detection/features/size_ratio.rb +26 -0
- data/lib/wikipedia/vandalism_detection/features/term_frequency.rb +25 -0
- data/lib/wikipedia/vandalism_detection/features/time_interval.rb +31 -0
- data/lib/wikipedia/vandalism_detection/features/time_of_day.rb +22 -0
- data/lib/wikipedia/vandalism_detection/features/upper_case_ratio.rb +24 -0
- data/lib/wikipedia/vandalism_detection/features/upper_case_words_ratio.rb +31 -0
- data/lib/wikipedia/vandalism_detection/features/upper_to_lower_case_ratio.rb +24 -0
- data/lib/wikipedia/vandalism_detection/features/user_reputation.rb +38 -0
- data/lib/wikipedia/vandalism_detection/features/vulgarism_frequency.rb +23 -0
- data/lib/wikipedia/vandalism_detection/features/vulgarism_impact.rb +22 -0
- data/lib/wikipedia/vandalism_detection/features/weekday.rb +21 -0
- data/lib/wikipedia/vandalism_detection/features/words_increment.rb +22 -0
- data/lib/wikipedia/vandalism_detection/instances.rb +130 -0
- data/lib/wikipedia/vandalism_detection/page.rb +88 -0
- data/lib/wikipedia/vandalism_detection/page_parser.rb +52 -0
- data/lib/wikipedia/vandalism_detection/revision.rb +69 -0
- data/lib/wikipedia/vandalism_detection/revision_parser.rb +43 -0
- data/lib/wikipedia/vandalism_detection/test_dataset.rb +367 -0
- data/lib/wikipedia/vandalism_detection/text.rb +18 -0
- data/lib/wikipedia/vandalism_detection/training_dataset.rb +303 -0
- data/lib/wikipedia/vandalism_detection/version.rb +5 -0
- data/lib/wikipedia/vandalism_detection/wikitext_extractor.rb +80 -0
- data/lib/wikipedia/vandalism_detection/word_lists.rb +19 -0
- data/lib/wikipedia/vandalism_detection/word_lists/bad.rb +12 -0
- data/lib/wikipedia/vandalism_detection/word_lists/biased.rb +21 -0
- data/lib/wikipedia/vandalism_detection/word_lists/emoticons.rb +22 -0
- data/lib/wikipedia/vandalism_detection/word_lists/markup.rb +12 -0
- data/lib/wikipedia/vandalism_detection/word_lists/pronouns.rb +15 -0
- data/lib/wikipedia/vandalism_detection/word_lists/sex.rb +12 -0
- data/lib/wikipedia/vandalism_detection/word_lists/vulgarism.rb +97 -0
- data/spec/factories/edit.rb +20 -0
- data/spec/factories/page.rb +13 -0
- data/spec/factories/revision.rb +51 -0
- data/spec/resources/config/config.yml +35 -0
- data/spec/resources/corpora/test/edits.csv +8 -0
- data/spec/resources/corpora/test/ground-truth.txt +3 -0
- data/spec/resources/corpora/test/revisions/part-1/326471754.txt +199 -0
- data/spec/resources/corpora/test/revisions/part-1/326873205.txt +137 -0
- data/spec/resources/corpora/test/revisions/part-1/328774035.txt +162 -0
- data/spec/resources/corpora/test/revisions/part-2/307084144.txt +137 -0
- data/spec/resources/corpora/test/revisions/part-2/326978767.txt +199 -0
- data/spec/resources/corpora/test/revisions/part-2/328774110.txt +162 -0
- data/spec/resources/corpora/training/annotations.csv +7 -0
- data/spec/resources/corpora/training/edits.csv +7 -0
- data/spec/resources/corpora/training/revisions/part-1/326471754.txt +199 -0
- data/spec/resources/corpora/training/revisions/part-1/326873205.txt +137 -0
- data/spec/resources/corpora/training/revisions/part-1/328774035.txt +162 -0
- data/spec/resources/corpora/training/revisions/part-2/307084144.txt +137 -0
- data/spec/resources/corpora/training/revisions/part-2/326978767.txt +199 -0
- data/spec/resources/corpora/training/revisions/part-2/328774110.txt +162 -0
- data/spec/resources/page_with_redirects.xml +85 -0
- data/spec/resources/redirect_page.xml +59 -0
- data/spec/resources/revision_simplified.xml +13 -0
- data/spec/resources/sample_revision.txt +137 -0
- data/spec/resources/sample_revision_clean_text.txt +1 -0
- data/spec/resources/sample_revision_plain_text.txt +183 -0
- data/spec/resources/vandalism_on_wikipedia.xml +234 -0
- data/spec/resources/vandalism_on_wikipedia_simplified.xml +119 -0
- data/spec/resources/wikipedia_tokens.txt +30 -0
- data/spec/spec_helper.rb +22 -0
- data/spec/support/macros/file_reading.rb +7 -0
- data/spec/support/macros/test_configuration.rb +71 -0
- data/spec/vandalism_detection/algorithms/kullback_leibler_divergence_spec.rb +36 -0
- data/spec/vandalism_detection/classifier_spec.rb +317 -0
- data/spec/vandalism_detection/configuration_spec.rb +517 -0
- data/spec/vandalism_detection/diff_spec.rb +40 -0
- data/spec/vandalism_detection/edit_spec.rb +137 -0
- data/spec/vandalism_detection/evaluator_spec.rb +671 -0
- data/spec/vandalism_detection/feature_calculator_spec.rb +128 -0
- data/spec/vandalism_detection/features/all_wordlists_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/all_wordlists_impact_spec.rb +58 -0
- data/spec/vandalism_detection/features/anonymity_previous_spec.rb +61 -0
- data/spec/vandalism_detection/features/anonymity_spec.rb +23 -0
- data/spec/vandalism_detection/features/article_size_spec.rb +35 -0
- data/spec/vandalism_detection/features/bad_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/bad_impact_spec.rb +59 -0
- data/spec/vandalism_detection/features/base_spec.rb +49 -0
- data/spec/vandalism_detection/features/biased_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/biased_impact_spec.rb +58 -0
- data/spec/vandalism_detection/features/blanking_spec.rb +38 -0
- data/spec/vandalism_detection/features/character_diversity_spec.rb +35 -0
- data/spec/vandalism_detection/features/character_sequence_spec.rb +37 -0
- data/spec/vandalism_detection/features/comment_bad_frequency_spec.rb +34 -0
- data/spec/vandalism_detection/features/comment_biased_frequency_spec.rb +34 -0
- data/spec/vandalism_detection/features/comment_length_spec.rb +27 -0
- data/spec/vandalism_detection/features/comment_markup_frequency_spec.rb +34 -0
- data/spec/vandalism_detection/features/comment_pronoun_frequency_spec.rb +34 -0
- data/spec/vandalism_detection/features/comment_sex_frequency_spec.rb +34 -0
- data/spec/vandalism_detection/features/comment_vulgarism_frequency_spec.rb +34 -0
- data/spec/vandalism_detection/features/compressibility_spec.rb +42 -0
- data/spec/vandalism_detection/features/contains_base_spec.rb +33 -0
- data/spec/vandalism_detection/features/copyedit_spec.rb +33 -0
- data/spec/vandalism_detection/features/digit_ratio_spec.rb +35 -0
- data/spec/vandalism_detection/features/edits_per_user_spec.rb +49 -0
- data/spec/vandalism_detection/features/emoticons_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/emoticons_impact_spec.rb +51 -0
- data/spec/vandalism_detection/features/frequency_base_spec.rb +26 -0
- data/spec/vandalism_detection/features/impact_base_spec.rb +41 -0
- data/spec/vandalism_detection/features/inserted_character_distribution_spec.rb +46 -0
- data/spec/vandalism_detection/features/inserted_external_links_spec.rb +35 -0
- data/spec/vandalism_detection/features/inserted_internal_links_spec.rb +35 -0
- data/spec/vandalism_detection/features/inserted_size_spec.rb +35 -0
- data/spec/vandalism_detection/features/inserted_words_spec.rb +35 -0
- data/spec/vandalism_detection/features/longest_word_spec.rb +35 -0
- data/spec/vandalism_detection/features/markup_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/markup_impact_spec.rb +59 -0
- data/spec/vandalism_detection/features/non_alphanumeric_ratio_spec.rb +35 -0
- data/spec/vandalism_detection/features/personal_life_spec.rb +26 -0
- data/spec/vandalism_detection/features/pronoun_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/pronoun_impact_spec.rb +59 -0
- data/spec/vandalism_detection/features/removed_all_wordlists_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/removed_bad_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/removed_biased_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/removed_character_distribution_spec.rb +46 -0
- data/spec/vandalism_detection/features/removed_emoticons_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/removed_markup_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/removed_pronoun_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/removed_sex_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/removed_size_spec.rb +35 -0
- data/spec/vandalism_detection/features/removed_vulgarism_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/removed_words_spec.rb +35 -0
- data/spec/vandalism_detection/features/replacement_similarity_spec.rb +44 -0
- data/spec/vandalism_detection/features/reverted_spec.rb +28 -0
- data/spec/vandalism_detection/features/revisions_character_distribution_spec.rb +46 -0
- data/spec/vandalism_detection/features/same_editor_spec.rb +60 -0
- data/spec/vandalism_detection/features/sex_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/sex_impact_spec.rb +59 -0
- data/spec/vandalism_detection/features/size_increment_spec.rb +35 -0
- data/spec/vandalism_detection/features/size_ratio_spec.rb +57 -0
- data/spec/vandalism_detection/features/term_frequency_spec.rb +38 -0
- data/spec/vandalism_detection/features/time_interval_spec.rb +50 -0
- data/spec/vandalism_detection/features/time_of_day_spec.rb +22 -0
- data/spec/vandalism_detection/features/upper_case_ratio_spec.rb +35 -0
- data/spec/vandalism_detection/features/upper_case_words_ratio_spec.rb +37 -0
- data/spec/vandalism_detection/features/upper_to_lower_case_ratio_spec.rb +35 -0
- data/spec/vandalism_detection/features/user_reputation_spec.rb +52 -0
- data/spec/vandalism_detection/features/vulgarism_frequency_spec.rb +36 -0
- data/spec/vandalism_detection/features/vulgarism_impact_spec.rb +58 -0
- data/spec/vandalism_detection/features/weekday_spec.rb +22 -0
- data/spec/vandalism_detection/features/words_increment_spec.rb +35 -0
- data/spec/vandalism_detection/instances_spec.rb +156 -0
- data/spec/vandalism_detection/page_parser_spec.rb +184 -0
- data/spec/vandalism_detection/page_spec.rb +135 -0
- data/spec/vandalism_detection/revision_parser_spec.rb +53 -0
- data/spec/vandalism_detection/revision_spec.rb +115 -0
- data/spec/vandalism_detection/test_dataset_spec.rb +231 -0
- data/spec/vandalism_detection/text_spec.rb +29 -0
- data/spec/vandalism_detection/training_dataset_spec.rb +264 -0
- data/spec/vandalism_detection/wikitext_extractor_spec.rb +72 -0
- data/spec/weka/classifiers/functions/lib_svm_spec.rb +38 -0
- data/spec/weka/classifiers/meta/one_class_classifier_spec.rb +76 -0
- data/spec/weka/classifiers/meta/real_ada_boost_spec.rb +40 -0
- data/spec/weka/classifiers/trees/balanced_random_forest_spec.rb +40 -0
- data/spec/weka/filters/supervised/instance/smote_spec.rb +6 -0
- data/wikipedia-vandalism_detection.gemspec +30 -0
- metadata +512 -0
@@ -0,0 +1,18 @@
|
|
1
|
+
require 'wikipedia/vandalism_detection/wikitext_extractor'
|
2
|
+
|
3
|
+
module Wikipedia
|
4
|
+
module VandalismDetection
|
5
|
+
class Text < String
|
6
|
+
|
7
|
+
def initialize(text = "")
|
8
|
+
super text.encode('UTF-8', 'binary', invalid: :replace, undef: :replace, replace: '')
|
9
|
+
end
|
10
|
+
|
11
|
+
# Extracts the plaintext from mediawiki markup and removes all line breaks & multiple spaces
|
12
|
+
# Return the cleaned plaintext.
|
13
|
+
def clean
|
14
|
+
@clean ||= WikitextExtractor.extract_clean self
|
15
|
+
end
|
16
|
+
end
|
17
|
+
end
|
18
|
+
end
|
@@ -0,0 +1,303 @@
|
|
1
|
+
require 'find'
|
2
|
+
require 'yaml'
|
3
|
+
require 'fileutils'
|
4
|
+
require 'active_support/core_ext/string'
|
5
|
+
require 'ruby-band'
|
6
|
+
|
7
|
+
require 'wikipedia/vandalism_detection/configuration'
|
8
|
+
require 'wikipedia/vandalism_detection/text'
|
9
|
+
require 'wikipedia/vandalism_detection/revision'
|
10
|
+
require 'wikipedia/vandalism_detection/edit'
|
11
|
+
require 'wikipedia/vandalism_detection/feature_calculator'
|
12
|
+
require 'wikipedia/vandalism_detection/instances'
|
13
|
+
require 'wikipedia/vandalism_detection/wikitext_extractor'
|
14
|
+
require 'weka/filters/supervised/instance/smote'
|
15
|
+
|
16
|
+
module Wikipedia
|
17
|
+
module VandalismDetection
|
18
|
+
|
19
|
+
# This class provides methods for getting and creating a training ARFF file from a configured training corpus.
|
20
|
+
class TrainingDataset
|
21
|
+
|
22
|
+
# Returns an instance dataset from the configured gold annotation file using the
|
23
|
+
# configured features from feature_calculator parameter.
|
24
|
+
def self.build
|
25
|
+
@config = Wikipedia::VandalismDetection.configuration
|
26
|
+
|
27
|
+
print "\ncreating training dataset..."
|
28
|
+
|
29
|
+
annotations_file = @config.training_corpus_annotations_file
|
30
|
+
raise AnnotationsFileNotConfiguredError unless annotations_file
|
31
|
+
|
32
|
+
annotations = CSV.parse(File.read(annotations_file), headers: true)
|
33
|
+
annotation_data = annotations.map { |row| { edit_id: row['editid'], class: row['class'] } }
|
34
|
+
|
35
|
+
output_directory = File.join(@config.output_base_directory, 'training')
|
36
|
+
FileUtils.mkdir_p(output_directory) unless Dir.exists?(output_directory)
|
37
|
+
FileUtils.mkdir_p(@config.output_base_directory) unless Dir.exists?(@config.output_base_directory)
|
38
|
+
|
39
|
+
# create feature file hash with io objects
|
40
|
+
feature_files = @config.features.inject({}) do |hash, feature_name|
|
41
|
+
file_name = "#{feature_name.gsub(' ', '_').downcase}.arff"
|
42
|
+
arff_file = File.join(output_directory, file_name)
|
43
|
+
|
44
|
+
unless File.exists?(arff_file)
|
45
|
+
dataset = Instances.empty_for_feature(feature_name)
|
46
|
+
dataset.to_ARFF(arff_file)
|
47
|
+
hash[feature_name] = File.open(arff_file, 'a')
|
48
|
+
end
|
49
|
+
|
50
|
+
hash
|
51
|
+
end
|
52
|
+
|
53
|
+
feature_calculator = FeatureCalculator.new
|
54
|
+
|
55
|
+
unless feature_files.empty?
|
56
|
+
processed_edits = 0
|
57
|
+
|
58
|
+
annotation_data.each do |row|
|
59
|
+
edit_id = row[:edit_id]
|
60
|
+
vandalism = row[:class]
|
61
|
+
edit = create_edit_from(edit_id)
|
62
|
+
|
63
|
+
feature_files.each do |feature_name, file|
|
64
|
+
value = feature_calculator.calculate_feature_for(edit, feature_name)
|
65
|
+
file.puts [value, vandalism].join(',')
|
66
|
+
end
|
67
|
+
|
68
|
+
processed_edits += 1
|
69
|
+
print_progress(processed_edits, @edits_csv.count, "computing training features")
|
70
|
+
end
|
71
|
+
|
72
|
+
# close all io objects
|
73
|
+
feature_files.each do |feature_name, file|
|
74
|
+
file.close
|
75
|
+
puts "'#{File.basename(file.path)}' saved to #{File.dirname(file.path)}"
|
76
|
+
end
|
77
|
+
end
|
78
|
+
|
79
|
+
dataset = merge_feature_arffs(@config.features, output_directory)
|
80
|
+
dataset.class_index = @config.features.count
|
81
|
+
dataset = replace_missing_values(dataset) if @config.replace_training_data_missing_values?
|
82
|
+
|
83
|
+
dataset
|
84
|
+
end
|
85
|
+
|
86
|
+
class << self
|
87
|
+
alias_method :instances, :build
|
88
|
+
end
|
89
|
+
|
90
|
+
# Returns the balanced training dataset (same number of vandalism & regular instances)
|
91
|
+
def self.balanced_instances
|
92
|
+
dataset = self.build
|
93
|
+
filter = Weka::Filters::Supervised::Instance::SpreadSubsample.new
|
94
|
+
|
95
|
+
#uniform distribution (remove majority instances)
|
96
|
+
filter.set do
|
97
|
+
data dataset
|
98
|
+
filter_options '-M 1'
|
99
|
+
end
|
100
|
+
|
101
|
+
filter.use
|
102
|
+
end
|
103
|
+
|
104
|
+
# Returns an oversampled training dataset.
|
105
|
+
# Oversampling options can be set by using e.g:
|
106
|
+
# percentage: 200
|
107
|
+
# undersampling: false
|
108
|
+
#
|
109
|
+
# For oversampling Weka SMOTE package is used.
|
110
|
+
# For SMOTE method see paper: http://arxiv.org/pdf/1106.1813.pdf
|
111
|
+
# Doc: http://weka.sourceforge.net/doc.packages/SMOTE/weka/filters/supervised/instance/SMOTE.html
|
112
|
+
def self.oversampled_instances(options = {})
|
113
|
+
config = Wikipedia::VandalismDetection.configuration
|
114
|
+
default_options = config.oversampling_options
|
115
|
+
options[:percentage] = default_options[:percentage] unless options[:percentage]
|
116
|
+
options[:undersampling] = default_options[:undersampling] unless options[:undersampling]
|
117
|
+
|
118
|
+
smote = Weka::Filters::Supervised::Instance::SMOTE.new
|
119
|
+
dataset = self.build
|
120
|
+
percentage = options[:percentage]
|
121
|
+
smote_options = "-P #{percentage.to_i}" if percentage
|
122
|
+
|
123
|
+
smote.set do
|
124
|
+
data dataset
|
125
|
+
filter_options smote_options if smote_options
|
126
|
+
end
|
127
|
+
|
128
|
+
undersampling = options[:undersampling] / 100.0
|
129
|
+
smote_dataset = smote.use
|
130
|
+
|
131
|
+
if undersampling > 0.0
|
132
|
+
subsample = Weka::Filters::Supervised::Instance::SpreadSubsample.new
|
133
|
+
|
134
|
+
# balance (remove majority instances)
|
135
|
+
subsample.set do
|
136
|
+
data smote_dataset
|
137
|
+
filter_options "-M #{undersampling}"
|
138
|
+
end
|
139
|
+
|
140
|
+
subsample.use
|
141
|
+
else
|
142
|
+
smote_dataset
|
143
|
+
end
|
144
|
+
end
|
145
|
+
|
146
|
+
def self.replace_missing_values(dataset)
|
147
|
+
puts "replacing missing values..."
|
148
|
+
filter = Weka::Filters::Unsupervised::Attribute::ReplaceMissingValues.new
|
149
|
+
filter.set { data dataset }
|
150
|
+
filter.use
|
151
|
+
end
|
152
|
+
|
153
|
+
# Saves and returns a file index hash of structure [file_name => full_path] for the given directory.
|
154
|
+
def self.create_corpus_file_index!
|
155
|
+
@config = Wikipedia::VandalismDetection.configuration
|
156
|
+
revisions_directory = @config.training_corpus_revisions_directory
|
157
|
+
|
158
|
+
raise RevisionsDirectoryNotConfiguredError unless revisions_directory
|
159
|
+
|
160
|
+
print "\ncreating file index..."
|
161
|
+
file_index = {}
|
162
|
+
|
163
|
+
Dir.open revisions_directory do |part_directories|
|
164
|
+
part_directories.each do |part_directory|
|
165
|
+
Dir.open "#{revisions_directory}/#{part_directory}" do |contents|
|
166
|
+
contents.each do |file|
|
167
|
+
path = "#{revisions_directory}/#{part_directory}/#{file}"
|
168
|
+
|
169
|
+
if File.file?(path) && (file =~ /\d+.txt/)
|
170
|
+
file_index[file] = path
|
171
|
+
print "\r processed #{file_index.count } files"
|
172
|
+
end
|
173
|
+
end
|
174
|
+
end
|
175
|
+
end
|
176
|
+
end
|
177
|
+
|
178
|
+
file = @config.training_output_index_file
|
179
|
+
dirname = File.dirname(file)
|
180
|
+
FileUtils.mkdir(dirname) unless Dir.exists?(dirname)
|
181
|
+
|
182
|
+
written = File.open(file, 'w') { |f| f.write(file_index.to_yaml) }
|
183
|
+
print "Index file saved to #{file}.\n" if written > 0
|
184
|
+
|
185
|
+
file_index
|
186
|
+
end
|
187
|
+
|
188
|
+
# Loads arff files of given features and merge them into one arff file.
|
189
|
+
# Returns the merged arff file.
|
190
|
+
def self.merge_feature_arffs(features, output_directory)
|
191
|
+
filter = Weka::Filters::Unsupervised::Attribute::Remove.new
|
192
|
+
merged_dataset = nil
|
193
|
+
|
194
|
+
features.map do |feature_name|
|
195
|
+
file_name = "#{feature_name.gsub(' ', '_').downcase}.arff"
|
196
|
+
arff_file = File.join(output_directory, file_name)
|
197
|
+
|
198
|
+
feature_dataset = Core::Parser.parse_ARFF(arff_file)
|
199
|
+
puts "using #{File.basename(arff_file)}"
|
200
|
+
|
201
|
+
if merged_dataset
|
202
|
+
filter.set do
|
203
|
+
data merged_dataset
|
204
|
+
filter_options '-R last'
|
205
|
+
end
|
206
|
+
|
207
|
+
merged_dataset = filter.use
|
208
|
+
merged_dataset = merged_dataset.merge_with(feature_dataset)
|
209
|
+
else
|
210
|
+
merged_dataset = feature_dataset
|
211
|
+
end
|
212
|
+
end
|
213
|
+
|
214
|
+
merged_dataset
|
215
|
+
end
|
216
|
+
|
217
|
+
# Creates a Wikipedia::Edit out of an annotation's edit id using files form config.yml
|
218
|
+
def self.create_edit_from(edit_id)
|
219
|
+
@file_index ||= load_corpus_file_index
|
220
|
+
edit_data = find_edits_data_for(edit_id)
|
221
|
+
|
222
|
+
old_revision_id = edit_data['oldrevisionid'].to_i
|
223
|
+
new_revision_id = edit_data['newrevisionid'].to_i
|
224
|
+
|
225
|
+
editor = edit_data['editor']
|
226
|
+
comment = edit_data['editcomment']
|
227
|
+
new_timestamp = edit_data['edittime']
|
228
|
+
page_id = edit_data['articleid']
|
229
|
+
page_title = edit_data['articletitle']
|
230
|
+
|
231
|
+
old_revision_file = @file_index["#{old_revision_id}.txt"]
|
232
|
+
new_revision_file = @file_index["#{new_revision_id}.txt"]
|
233
|
+
|
234
|
+
raise(RevisionFileNotFound, "Old revision file #{old_revision_file} not found") unless \
|
235
|
+
File.exist?(old_revision_file)
|
236
|
+
raise(RevisionFileNotFound, "New revision file #{new_revision_file} not found") unless \
|
237
|
+
File.exist?(new_revision_file)
|
238
|
+
|
239
|
+
old_revision_text = File.read(old_revision_file)
|
240
|
+
new_revision_text = File.read(new_revision_file)
|
241
|
+
|
242
|
+
old_revision = Revision.new
|
243
|
+
old_revision.id = old_revision_id
|
244
|
+
old_revision.text = Text.new(old_revision_text)
|
245
|
+
|
246
|
+
new_revision = Revision.new
|
247
|
+
new_revision.id = new_revision_id
|
248
|
+
new_revision.text = Text.new(new_revision_text)
|
249
|
+
new_revision.parent_id = old_revision_id
|
250
|
+
new_revision.comment = Text.new(comment)
|
251
|
+
new_revision.contributor = editor
|
252
|
+
new_revision.timestamp = new_timestamp
|
253
|
+
|
254
|
+
page = Page.new
|
255
|
+
page.id = page_id
|
256
|
+
page.title = page_title
|
257
|
+
|
258
|
+
Edit.new(old_revision, new_revision, page: page)
|
259
|
+
end
|
260
|
+
|
261
|
+
# Gets or creates the corpus index file, which holds a hash of revision files name and their path
|
262
|
+
# in the article revisions directory.
|
263
|
+
def self.load_corpus_file_index
|
264
|
+
index_file = @config.training_output_index_file
|
265
|
+
|
266
|
+
if File.exists? index_file
|
267
|
+
puts "\n(Using #{index_file})\n"
|
268
|
+
YAML.load_file index_file
|
269
|
+
else
|
270
|
+
create_corpus_file_index!
|
271
|
+
end
|
272
|
+
end
|
273
|
+
|
274
|
+
# Returns the line array of the edits.csv file with given edit id.
|
275
|
+
def self.find_edits_data_for(edit_id)
|
276
|
+
edits_file = Wikipedia::VandalismDetection.configuration.training_corpus_edits_file
|
277
|
+
raise EditsFileNotConfiguredError unless edits_file
|
278
|
+
|
279
|
+
@edits_file_content ||= File.read(edits_file)
|
280
|
+
@edits_csv ||= CSV.parse(@edits_file_content, headers: true)
|
281
|
+
|
282
|
+
edit_data = @edits_csv.find { |row| row['editid'] == edit_id }
|
283
|
+
raise "Edit data for edit id #{edit_id} not found in #{ File.basename(edits_file) }." unless edit_data
|
284
|
+
|
285
|
+
edit_data
|
286
|
+
end
|
287
|
+
|
288
|
+
# Prints the progress to the $stdout
|
289
|
+
def self.print_progress(processed_count, total_count, message)
|
290
|
+
processed_absolute = "#{processed_count}/#{total_count}"
|
291
|
+
processed_percentage = "%0.2f%" % ((processed_count * 100.00) / total_count).round(2)
|
292
|
+
print "\r#{message}... #{processed_absolute} | #{processed_percentage}"
|
293
|
+
end
|
294
|
+
|
295
|
+
private_class_method :create_edit_from,
|
296
|
+
:merge_feature_arffs,
|
297
|
+
:print_progress,
|
298
|
+
:find_edits_data_for,
|
299
|
+
:load_corpus_file_index,
|
300
|
+
:replace_missing_values
|
301
|
+
end
|
302
|
+
end
|
303
|
+
end
|
@@ -0,0 +1,80 @@
|
|
1
|
+
# The WikitextExtractor imports the WikitextExtractor class from the sweble-wikitext-extractor.jar
|
2
|
+
# The sweble-wikitext-extractor.jar is a custom Java project which uses the Sweble wikitext parser to extract
|
3
|
+
# plaintext out of wikimarkup text.
|
4
|
+
#
|
5
|
+
# The Sweble WikitextExtractor currently depends on the swc-engine -v1.1.0 with dependencies,
|
6
|
+
# see: http://sweble.org/downloads/swc-devel/master-latest/ to download it.
|
7
|
+
#
|
8
|
+
# The Java source code can be found on:
|
9
|
+
# webis.uni-weimar.de:/srv/cvsroot/code-in-progress/wikipedia-vandalism-detection/sweble-wikitext-extractor
|
10
|
+
#
|
11
|
+
# @author Paul Götze <paul.christoph.goetze@gmail.com>
|
12
|
+
module Wikipedia
|
13
|
+
module VandalismDetection
|
14
|
+
|
15
|
+
require 'java'
|
16
|
+
require 'java/swc-engine-1.1.0-jar-with-dependencies.jar'
|
17
|
+
require 'java/sweble-wikitext-extractor.jar'
|
18
|
+
|
19
|
+
java_import 'de.webis.sweble.WikitextExtractor'
|
20
|
+
|
21
|
+
class WikitextExtractionError < StandardError; end
|
22
|
+
|
23
|
+
# This class wrapps the de.webis.sweble.WikitextExtractor Java class and provides methods to extract plaintext
|
24
|
+
# from wiki markup text both space preserving and cleaned without line breaks and whitespace.
|
25
|
+
#
|
26
|
+
# @author Paul Götze <paul.christoph.goetze@gmail.com>
|
27
|
+
class WikitextExtractor
|
28
|
+
|
29
|
+
REDIRECT = '#REDIRECT'
|
30
|
+
|
31
|
+
# Returns the extracted text from the given wiki markup preserving spacing with added section numbers.
|
32
|
+
def self.extract(wiki_text)
|
33
|
+
begin
|
34
|
+
wiki_text = wiki_text.encode('UTF-8', 'binary', invalid: :replace, undef: :replace, replace: '')
|
35
|
+
wiki_text = wiki_text.gsub(REDIRECT, '')
|
36
|
+
|
37
|
+
WikitextExtractor.new.extract(wiki_text)
|
38
|
+
rescue => exception
|
39
|
+
raise WikitextExtractionError, "Wikitext extraction failed: \n#{exception.message}", caller
|
40
|
+
end
|
41
|
+
end
|
42
|
+
|
43
|
+
# Returns the cleaned extracted text from the given wiki markup.
|
44
|
+
# Cleaned means a single string without breaks, multiple spaces and section numbers.
|
45
|
+
def self.extract_clean(wiki_text)
|
46
|
+
wiki_text = extract wiki_text
|
47
|
+
|
48
|
+
wiki_text = remove_section_numbering_from wiki_text
|
49
|
+
wiki_text = remove_line_breaks_from wiki_text
|
50
|
+
wiki_text = remove_uris_from wiki_text
|
51
|
+
wiki_text = remove_special_signes_from wiki_text
|
52
|
+
wiki_text = remove_multiple_spaces_from wiki_text
|
53
|
+
wiki_text = wiki_text.strip
|
54
|
+
end
|
55
|
+
|
56
|
+
private
|
57
|
+
|
58
|
+
# removes 1., 1.1., 2.3.4. etc. at the beginning of a line
|
59
|
+
def self.remove_section_numbering_from(text)
|
60
|
+
text.gsub /^(\d\.)+/, ''
|
61
|
+
end
|
62
|
+
|
63
|
+
def self.remove_line_breaks_from(text)
|
64
|
+
text.gsub /\n+/, ' '
|
65
|
+
end
|
66
|
+
|
67
|
+
def self.remove_multiple_spaces_from(text)
|
68
|
+
text.gsub /\s+/, ' '
|
69
|
+
end
|
70
|
+
|
71
|
+
def self.remove_uris_from(text)
|
72
|
+
text.gsub /(https?|ftp)\s?:\s?\/\/[^\s\/$.?#].[^\s]*/i, ''
|
73
|
+
end
|
74
|
+
|
75
|
+
def self.remove_special_signes_from(text)
|
76
|
+
text.gsub /\[\]\{\}\|\=/, ' '
|
77
|
+
end
|
78
|
+
end
|
79
|
+
end
|
80
|
+
end
|
@@ -0,0 +1,19 @@
|
|
1
|
+
require 'wikipedia/vandalism_detection/word_lists/bad'
|
2
|
+
require 'wikipedia/vandalism_detection/word_lists/biased'
|
3
|
+
require 'wikipedia/vandalism_detection/word_lists/pronouns'
|
4
|
+
require 'wikipedia/vandalism_detection/word_lists/sex'
|
5
|
+
require 'wikipedia/vandalism_detection/word_lists/vulgarism'
|
6
|
+
require 'wikipedia/vandalism_detection/word_lists/markup'
|
7
|
+
|
8
|
+
module Wikipedia
|
9
|
+
module VandalismDetection
|
10
|
+
module WordLists
|
11
|
+
|
12
|
+
# Returns an array of all wordlist words
|
13
|
+
def self.all
|
14
|
+
[*BAD, *BIASED, *PRONOUNS, *SEX, *VULGARISM].uniq!
|
15
|
+
end
|
16
|
+
|
17
|
+
end
|
18
|
+
end
|
19
|
+
end
|
@@ -0,0 +1,12 @@
|
|
1
|
+
module Wikipedia
|
2
|
+
module VandalismDetection
|
3
|
+
module WordLists
|
4
|
+
|
5
|
+
BAD =
|
6
|
+
[
|
7
|
+
:"666", :da, :dont, :dosent, :whatever, :guy, :hi, :nazi, :sup, :guise, :loser, :thats, :ugly, :wanna,
|
8
|
+
:whats, :wont, :gotta, :bloody, :fart, :pot, :prick, :stink, :smells, :smelly, :alot, :dunno, :gotcha
|
9
|
+
]
|
10
|
+
end
|
11
|
+
end
|
12
|
+
end
|
@@ -0,0 +1,21 @@
|
|
1
|
+
module Wikipedia
|
2
|
+
module VandalismDetection
|
3
|
+
module WordLists
|
4
|
+
|
5
|
+
BIASED =
|
6
|
+
[
|
7
|
+
:acclaimed, :amazing, :astonishing, :authoritative, :beautiful, :best, :brilliant, :canonical,
|
8
|
+
:celebrated, :charismatic, :classic, :"cutting-edge", :defining, :definitive, :eminent, :enigma,
|
9
|
+
:exciting, :extraordinary, :fabulous, :famous, :infamous, :fantastic, :fully, :genius, :global,
|
10
|
+
:great, :greatest, :iconic, :immensely, :impactful, :incendiary, :indisputable, :influential,
|
11
|
+
:innovative, :inspired, :intriguing, :leader, :leading, :legendary, :major, :masterly, :mature,
|
12
|
+
:memorable, :notable, :outstanding, :pioneer, :popular, :prestigious, :really, :remarkable,
|
13
|
+
:renowned, :respected, :seminal, :significant, :skillful, :solution, :"single-handedly", :staunch,
|
14
|
+
:talented, :most, :top, :transcendent, :undoubtedly, :unique, :visionary, :virtually,
|
15
|
+
:virtuoso, :"well-known", :"well-established", :"world-class", :worst, :coolest, :super, :probably,
|
16
|
+
:hate, :ugly, :fat, :lame, :weird, :strange, :everyone, :cares, :boring, :boreing, :ever, :huge,
|
17
|
+
:like, :idiotic, :absolute, :total, :totally
|
18
|
+
]
|
19
|
+
end
|
20
|
+
end
|
21
|
+
end
|