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 +1,69 @@
|
|
1
|
+
require_relative 'text'
|
2
|
+
require 'zlib'
|
3
|
+
|
4
|
+
module Wikipedia
|
5
|
+
module VandalismDetection
|
6
|
+
class Revision
|
7
|
+
|
8
|
+
START_TAG = '<revision>'.freeze
|
9
|
+
END_TAG = '</revision>'.freeze
|
10
|
+
REDIRECT_PATTERN = /#REDIRECT\s+\[\[.*?\]\]/
|
11
|
+
|
12
|
+
attr_accessor :id,
|
13
|
+
:parent_id,
|
14
|
+
:timestamp,
|
15
|
+
:comment,
|
16
|
+
:contributor_username,
|
17
|
+
:sha1
|
18
|
+
|
19
|
+
attr_reader :contributor_id,
|
20
|
+
:contributor_ip
|
21
|
+
|
22
|
+
def initialize
|
23
|
+
@text = Zlib::Deflate.deflate('')
|
24
|
+
@comment = Text.new
|
25
|
+
end
|
26
|
+
|
27
|
+
def contributor=(contributor)
|
28
|
+
if ip_v4? contributor
|
29
|
+
@contributor_ip = contributor
|
30
|
+
else
|
31
|
+
@contributor_id = contributor
|
32
|
+
end
|
33
|
+
end
|
34
|
+
|
35
|
+
def contributor
|
36
|
+
@contributor_id || @contributor_ip
|
37
|
+
end
|
38
|
+
|
39
|
+
def anonymous_contributor?
|
40
|
+
!@contributor_ip.nil?
|
41
|
+
end
|
42
|
+
|
43
|
+
def redirect?
|
44
|
+
!!(text =~ REDIRECT_PATTERN)
|
45
|
+
end
|
46
|
+
|
47
|
+
# Compresses text when set
|
48
|
+
def text=(text)
|
49
|
+
# remove invalid utf-8 byte sequences
|
50
|
+
text.encode!('UTF-16', 'UTF-8', invalid: :replace, replace: '')
|
51
|
+
text.encode!('UTF-8', 'UTF-16')
|
52
|
+
@text = Zlib::Deflate.deflate(text)
|
53
|
+
end
|
54
|
+
|
55
|
+
# Decompresses text when called
|
56
|
+
def text
|
57
|
+
Text.new(Zlib::Inflate.inflate(@text).force_encoding('utf-8'))
|
58
|
+
end
|
59
|
+
|
60
|
+
private
|
61
|
+
|
62
|
+
# Returns whether the given value is an IPv4.
|
63
|
+
def ip_v4?(value)
|
64
|
+
!!value.to_s.match(/(\d+)\.(\d+)\.(\d+)\.(\d+)/)
|
65
|
+
end
|
66
|
+
|
67
|
+
end
|
68
|
+
end
|
69
|
+
end
|
@@ -0,0 +1,43 @@
|
|
1
|
+
# encoding: utf-8
|
2
|
+
|
3
|
+
require 'nokogiri'
|
4
|
+
require 'wikipedia/vandalism_detection/revision'
|
5
|
+
|
6
|
+
module Wikipedia
|
7
|
+
module VandalismDetection
|
8
|
+
class RevisionParser
|
9
|
+
|
10
|
+
# Parses an xml string and returns a Wikipedia::VandalismDetection::Revision.
|
11
|
+
def parse(xml, options = {})
|
12
|
+
document = Nokogiri::XML(xml, nil, 'UTF-8').xpath('//revision')
|
13
|
+
|
14
|
+
revision = Revision.new
|
15
|
+
|
16
|
+
properties = options[:only] || [:id, :parent_id, :timestamp, :contributor, :comment, :text, :sha1]
|
17
|
+
|
18
|
+
revision.id = document.xpath('id').inner_text if properties.include?(:id)
|
19
|
+
revision.timestamp = document.xpath('timestamp').inner_text if properties.include?(:timestamp)
|
20
|
+
revision.comment = Text.new(document.xpath('comment').inner_text) if properties.include?(:comment)
|
21
|
+
revision.text = Text.new(document.xpath('text').inner_text) if properties.include?(:text)
|
22
|
+
revision.sha1 = document.xpath('sha1').inner_text if properties.include?(:sha1)
|
23
|
+
|
24
|
+
if properties.include?(:contributor)
|
25
|
+
contributor_id_node = document.xpath('contributor/id')
|
26
|
+
contributor_ip_node = document.xpath('contributor/ip')
|
27
|
+
contributor_username_node = document.xpath('contributor/username')
|
28
|
+
|
29
|
+
revision.contributor = contributor_id_node.inner_text unless contributor_id_node.empty?
|
30
|
+
revision.contributor = contributor_ip_node.inner_text unless contributor_ip_node.empty?
|
31
|
+
revision.contributor_username = contributor_username_node.inner_text unless contributor_username_node.empty?
|
32
|
+
end
|
33
|
+
|
34
|
+
if properties.include?(:parent_id)
|
35
|
+
parent_id_node = document.xpath('parentid')
|
36
|
+
revision.parent_id = parent_id_node.inner_text unless parent_id_node.empty?
|
37
|
+
end
|
38
|
+
|
39
|
+
revision
|
40
|
+
end
|
41
|
+
end
|
42
|
+
end
|
43
|
+
end
|
@@ -0,0 +1,367 @@
|
|
1
|
+
require 'find'
|
2
|
+
require 'yaml'
|
3
|
+
require 'fileutils'
|
4
|
+
require 'csv'
|
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
|
+
|
15
|
+
module Wikipedia
|
16
|
+
module VandalismDetection
|
17
|
+
|
18
|
+
# This class provides methods for getting and creating a test ARFF file from a configured test corpus.
|
19
|
+
class TestDataset
|
20
|
+
|
21
|
+
# Returns an instance dataset from the configured gold annotation file using the
|
22
|
+
# configured features from feature_calculator parameter.
|
23
|
+
def self.build
|
24
|
+
@config = Wikipedia::VandalismDetection.configuration
|
25
|
+
print "\ncreating test dataset..."
|
26
|
+
|
27
|
+
edits_file = @config.test_corpus_edits_file
|
28
|
+
raise EditsFileNotConfiguredError unless edits_file
|
29
|
+
|
30
|
+
edits = CSV.parse(File.read(edits_file), headers: true)
|
31
|
+
|
32
|
+
output_directory = File.join(@config.output_base_directory, 'test')
|
33
|
+
FileUtils.mkdir_p(output_directory) unless Dir.exists?(output_directory)
|
34
|
+
FileUtils.mkdir_p(@config.output_base_directory) unless Dir.exists?(@config.output_base_directory)
|
35
|
+
|
36
|
+
# create feature file hash with io objects
|
37
|
+
feature_files = @config.features.inject({}) do |hash, feature_name|
|
38
|
+
file_name = "#{feature_name.gsub(' ', '_').downcase}.arff"
|
39
|
+
arff_file = File.join(output_directory, file_name)
|
40
|
+
|
41
|
+
unless File.exists?(arff_file)
|
42
|
+
dataset = Instances.empty_for_test_feature(feature_name)
|
43
|
+
dataset.to_ARFF(arff_file)
|
44
|
+
hash[feature_name] = File.open(arff_file, 'a')
|
45
|
+
end
|
46
|
+
|
47
|
+
hash
|
48
|
+
end
|
49
|
+
|
50
|
+
feature_calculator = FeatureCalculator.new
|
51
|
+
|
52
|
+
unless feature_files.empty?
|
53
|
+
processed_edits = 0
|
54
|
+
edits_count = edits.count
|
55
|
+
|
56
|
+
edits.each do |edit_data|
|
57
|
+
old_revision_id = edit_data['oldrevisionid']
|
58
|
+
new_revision_id = edit_data['newrevisionid']
|
59
|
+
|
60
|
+
processed_edits += 1
|
61
|
+
print_progress(processed_edits, edits_count, "computing test features")
|
62
|
+
|
63
|
+
next unless (annotated_revision?(old_revision_id) && annotated_revision?(new_revision_id))
|
64
|
+
edit = create_edit_from(edit_data)
|
65
|
+
|
66
|
+
feature_files.each do |feature_name, file|
|
67
|
+
value = feature_calculator.calculate_feature_for(edit, feature_name)
|
68
|
+
file.puts [value, old_revision_id, new_revision_id].join(',')
|
69
|
+
end
|
70
|
+
end
|
71
|
+
|
72
|
+
# close all io objects
|
73
|
+
feature_files.each do |feature_name, file|
|
74
|
+
file.close
|
75
|
+
puts "\n'#{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 = normalize(dataset) if @config.classifier_type.match('Functions::LibSVM')
|
81
|
+
dataset
|
82
|
+
end
|
83
|
+
|
84
|
+
class << self
|
85
|
+
alias_method :instances, :build
|
86
|
+
end
|
87
|
+
|
88
|
+
# Saves and returns the dataset as ARFF file.
|
89
|
+
# As test data the configured data corpus from /config/config.yml is used.
|
90
|
+
def self.build!
|
91
|
+
@config = Wikipedia::VandalismDetection.configuration
|
92
|
+
|
93
|
+
dataset = self.instances
|
94
|
+
output_file = @config.test_output_arff_file
|
95
|
+
dataset.to_ARFF(output_file)
|
96
|
+
puts "\n'#{File.basename(output_file)}' saved to #{File.dirname(output_file)}"
|
97
|
+
|
98
|
+
dataset
|
99
|
+
end
|
100
|
+
|
101
|
+
# Loads arff files of given features and merge them into one arff file.
|
102
|
+
# Returns the merged arff file.
|
103
|
+
def self.merge_feature_arffs(features, output_directory)
|
104
|
+
filter = Weka::Filters::Unsupervised::Attribute::Remove.new
|
105
|
+
merged_dataset = nil
|
106
|
+
|
107
|
+
features.each do |feature_name|
|
108
|
+
file_name = "#{feature_name.gsub(' ', '_').downcase}.arff"
|
109
|
+
arff_file = File.join(output_directory, file_name)
|
110
|
+
|
111
|
+
feature_dataset = Core::Parser.parse_ARFF(arff_file)
|
112
|
+
print "."
|
113
|
+
|
114
|
+
if merged_dataset
|
115
|
+
2.times do
|
116
|
+
filter.set do
|
117
|
+
data merged_dataset
|
118
|
+
filter_options '-R last'
|
119
|
+
end
|
120
|
+
|
121
|
+
merged_dataset = filter.use
|
122
|
+
end
|
123
|
+
|
124
|
+
merged_dataset = merged_dataset.merge_with(feature_dataset)
|
125
|
+
else
|
126
|
+
merged_dataset = feature_dataset
|
127
|
+
end
|
128
|
+
end
|
129
|
+
|
130
|
+
add_ground_truth_class_to(merged_dataset)
|
131
|
+
end
|
132
|
+
|
133
|
+
# Adds the ground truth class attribute and values to the given dataset and returns the merged
|
134
|
+
def self.add_ground_truth_class_to(dataset)
|
135
|
+
config = Wikipedia::VandalismDetection.configuration
|
136
|
+
|
137
|
+
arff_file = File.join(config.output_base_directory, 'test', 'class.arff')
|
138
|
+
class_dataset = Instances.empty_for_test_class
|
139
|
+
|
140
|
+
if File.exists?(arff_file)
|
141
|
+
class_dataset = Core::Parser.parse_ARFF(arff_file)
|
142
|
+
else
|
143
|
+
ground_truth_file_path = config.test_corpus_ground_truth_file
|
144
|
+
ground_truth = ground_truth_hash(ground_truth_file_path)
|
145
|
+
|
146
|
+
dataset.to_a2d.each do |instance|
|
147
|
+
old_revision_id = instance[-2].to_i
|
148
|
+
new_revision_id = instance[-1].to_i
|
149
|
+
key = :"#{old_revision_id}-#{new_revision_id}"
|
150
|
+
|
151
|
+
if ground_truth.has_key?(key)
|
152
|
+
class_value = Instances::CLASSES[Instances::CLASSES_SHORT.key(ground_truth[key][:class])]
|
153
|
+
class_value ? class_dataset.add_instance([class_value]) : class_dataset.add(Core::Type::DenseInstance.new(1))
|
154
|
+
else
|
155
|
+
class_dataset.add(Core::Type::DenseInstance.new(1)) # missing
|
156
|
+
end
|
157
|
+
end
|
158
|
+
|
159
|
+
class_dataset.to_ARFF(arff_file)
|
160
|
+
puts "saved #{File.basename(arff_file)} to #{File.dirname(arff_file)}"
|
161
|
+
end
|
162
|
+
|
163
|
+
dataset.merge_with(class_dataset)
|
164
|
+
end
|
165
|
+
|
166
|
+
# Returns a hash for classification data from given ground truth file
|
167
|
+
def self.ground_truth_hash(ground_truth_file)
|
168
|
+
file = File.read(ground_truth_file)
|
169
|
+
ground_truth_samples = file.lines.to_a
|
170
|
+
|
171
|
+
ground_truth = {}
|
172
|
+
|
173
|
+
ground_truth_samples.each do |line|
|
174
|
+
line_parts = line.split(' ')
|
175
|
+
|
176
|
+
old_revision_id = line_parts[0].to_i
|
177
|
+
new_revision_id = line_parts[1].to_i
|
178
|
+
class_short = line_parts[2]
|
179
|
+
|
180
|
+
ground_truth[:"#{old_revision_id}-#{new_revision_id}"] = {
|
181
|
+
old_revision_id: old_revision_id,
|
182
|
+
new_revision_id: new_revision_id,
|
183
|
+
class: class_short
|
184
|
+
}
|
185
|
+
end
|
186
|
+
|
187
|
+
ground_truth
|
188
|
+
end
|
189
|
+
|
190
|
+
# Saves and returns a file index hash of structure [file_name => full_path] for the given directory.
|
191
|
+
def self.create_corpus_file_index!
|
192
|
+
@config = Wikipedia::VandalismDetection.configuration
|
193
|
+
revisions_directory = @config.test_corpus_revisions_directory
|
194
|
+
|
195
|
+
raise RevisionsDirectoryNotConfiguredError unless revisions_directory
|
196
|
+
|
197
|
+
print "\nCreating test corpus index file..."
|
198
|
+
file_index = {}
|
199
|
+
|
200
|
+
Dir.open revisions_directory do |part_directories|
|
201
|
+
part_directories.each do |part_directory|
|
202
|
+
Dir.open "#{revisions_directory}/#{part_directory}" do |contents|
|
203
|
+
contents.each do |file|
|
204
|
+
path = "#{revisions_directory}/#{part_directory}/#{file}"
|
205
|
+
|
206
|
+
if File.file?(path) && (file =~ /\d+.txt/) && annotated_revision?(file)
|
207
|
+
file_index[file] = path
|
208
|
+
print "\r processed #{file_index.count } files"
|
209
|
+
end
|
210
|
+
end
|
211
|
+
end
|
212
|
+
end
|
213
|
+
end
|
214
|
+
|
215
|
+
file = @config.test_output_index_file
|
216
|
+
dirname = File.dirname(file)
|
217
|
+
FileUtils.mkdir(dirname) unless Dir.exists?(dirname)
|
218
|
+
|
219
|
+
written = File.open(file, 'w') { |f| f.write(file_index.to_yaml) }
|
220
|
+
print "\nSaved test corpus index file to #{file}.\n" if written > 0
|
221
|
+
|
222
|
+
file_index
|
223
|
+
end
|
224
|
+
|
225
|
+
# Returns whether the given revision is annotated in the configured gold annotation file.
|
226
|
+
def self.annotated_revision?(revision_file_or_id)
|
227
|
+
@annotated_revisions ||= annotated_revisions
|
228
|
+
|
229
|
+
revision_id = revision_file_or_id.to_s.gsub('.txt', '')
|
230
|
+
@annotated_revisions[revision_id.to_sym]
|
231
|
+
end
|
232
|
+
|
233
|
+
# Returns a Hash with the used revision ids from edits_file.
|
234
|
+
def self.annotated_revisions
|
235
|
+
annotations_file = @config.test_corpus_ground_truth_file
|
236
|
+
annotations = File.read(annotations_file).lines
|
237
|
+
|
238
|
+
annotated_revisions = {}
|
239
|
+
|
240
|
+
annotations.each do |annotation|
|
241
|
+
data = annotation.split(' ')
|
242
|
+
|
243
|
+
annotated_revisions[data[0].to_sym] = true
|
244
|
+
annotated_revisions[data[1].to_sym] = true
|
245
|
+
end
|
246
|
+
|
247
|
+
@annotated_revisions ||= annotated_revisions
|
248
|
+
end
|
249
|
+
|
250
|
+
# Removes all instances with missing attributes
|
251
|
+
def self.remove_missing(dataset)
|
252
|
+
dataset.each_column do |attribute|
|
253
|
+
dataset.delete_with_missing(attribute)
|
254
|
+
end
|
255
|
+
|
256
|
+
dataset
|
257
|
+
end
|
258
|
+
|
259
|
+
# Returns the normalized dataset (important for lib svm one class classification)
|
260
|
+
def self.normalize(dataset)
|
261
|
+
remove_filter = Weka::Filters::Unsupervised::Attribute::Remove.new
|
262
|
+
remove_filter.data(dataset)
|
263
|
+
remove_filter.filter_options("-V -R 1-#{@config.features.count}")
|
264
|
+
numerics_dataset = remove_filter.use
|
265
|
+
|
266
|
+
remove_filter.filter_options("-R 1-#{@config.features.count}")
|
267
|
+
non_numerics_dataset = remove_filter.use
|
268
|
+
|
269
|
+
normalize_filter = Weka::Filters::Unsupervised::Attribute::Normalize.new
|
270
|
+
normalize_filter.data(numerics_dataset)
|
271
|
+
normalized_dataset = normalize_filter.use
|
272
|
+
|
273
|
+
normalized_dataset.merge_with non_numerics_dataset
|
274
|
+
end
|
275
|
+
|
276
|
+
# Returns the Edit with the given revision ids.
|
277
|
+
# Test corpus is searched for the revisions' data.
|
278
|
+
def self.edit(old_revision_id, new_revision_id)
|
279
|
+
@config = Wikipedia::VandalismDetection.configuration
|
280
|
+
edits_file = @config.test_corpus_edits_file
|
281
|
+
raise EditsFileNotConfiguredError unless edits_file
|
282
|
+
|
283
|
+
@edits_csv ||= CSV.parse(File.read(edits_file), headers: true)
|
284
|
+
|
285
|
+
edit_data = @edits_csv.find do |row|
|
286
|
+
row['oldrevisionid'] == old_revision_id && row['newrevisionid'] == new_revision_id
|
287
|
+
end
|
288
|
+
|
289
|
+
if (annotated_revision?(old_revision_id) && annotated_revision?(new_revision_id)) && edit_data
|
290
|
+
create_edit_from(edit_data)
|
291
|
+
end
|
292
|
+
end
|
293
|
+
|
294
|
+
# Creates a Wikipedia::Edit out of an edit's data from edit_file configured in config.yml
|
295
|
+
def self.create_edit_from(edit_data)
|
296
|
+
@file_index ||= load_corpus_file_index
|
297
|
+
|
298
|
+
old_revision_id = edit_data['oldrevisionid'].to_i
|
299
|
+
new_revision_id = edit_data['newrevisionid'].to_i
|
300
|
+
|
301
|
+
editor = edit_data['editor']
|
302
|
+
comment = edit_data['editcomment']
|
303
|
+
new_timestamp = edit_data['edittime']
|
304
|
+
page_id = edit_data['articleid']
|
305
|
+
page_title = edit_data['articletitle']
|
306
|
+
|
307
|
+
old_revision_file = @file_index["#{old_revision_id}.txt"]
|
308
|
+
new_revision_file = @file_index["#{new_revision_id}.txt"]
|
309
|
+
|
310
|
+
raise(RevisionFileNotFound, "Old revision file #{old_revision_file} not found") unless \
|
311
|
+
File.exist?(old_revision_file)
|
312
|
+
raise(RevisionFileNotFound, "New revision file #{new_revision_file} not found") unless \
|
313
|
+
File.exist?(new_revision_file)
|
314
|
+
|
315
|
+
old_revision_text = File.read(old_revision_file)
|
316
|
+
new_revision_text = File.read(new_revision_file)
|
317
|
+
|
318
|
+
old_revision = Revision.new
|
319
|
+
old_revision.id = old_revision_id
|
320
|
+
old_revision.text = Text.new(old_revision_text)
|
321
|
+
|
322
|
+
new_revision = Revision.new
|
323
|
+
new_revision.id = new_revision_id
|
324
|
+
new_revision.text = Text.new(new_revision_text)
|
325
|
+
new_revision.parent_id = old_revision_id
|
326
|
+
new_revision.comment = Text.new(comment)
|
327
|
+
new_revision.contributor = editor
|
328
|
+
new_revision.timestamp = new_timestamp
|
329
|
+
|
330
|
+
page = Page.new
|
331
|
+
page.id = page_id
|
332
|
+
page.title = page_title
|
333
|
+
|
334
|
+
Edit.new(old_revision, new_revision, page: page)
|
335
|
+
end
|
336
|
+
|
337
|
+
# Gets or creates the corpus index file, which holds a hash of revision files name and their path
|
338
|
+
# in the article revisions directory.
|
339
|
+
def self.load_corpus_file_index
|
340
|
+
index_file = @config.test_output_index_file
|
341
|
+
|
342
|
+
if File.exists? index_file
|
343
|
+
puts " (Using #{index_file}) \n"
|
344
|
+
YAML.load_file index_file
|
345
|
+
else
|
346
|
+
create_corpus_file_index!
|
347
|
+
end
|
348
|
+
end
|
349
|
+
|
350
|
+
# Prints the progress to the $stdout
|
351
|
+
def self.print_progress(processed_count, total_count, message)
|
352
|
+
processed_absolute = "#{processed_count}/#{total_count}"
|
353
|
+
processed_percentage = "%0.2f%" % ((processed_count * 100.00) / total_count).round(2)
|
354
|
+
print "\r#{message}... #{processed_absolute} | #{processed_percentage}"
|
355
|
+
end
|
356
|
+
|
357
|
+
private_class_method :annotated_revision?,
|
358
|
+
:annotated_revisions,
|
359
|
+
:remove_missing,
|
360
|
+
:create_edit_from,
|
361
|
+
:load_corpus_file_index,
|
362
|
+
:print_progress,
|
363
|
+
:merge_feature_arffs,
|
364
|
+
:add_ground_truth_class_to
|
365
|
+
end
|
366
|
+
end
|
367
|
+
end
|