Merge pull request #28900 from code-dot-org/custom-language-filter

Language-specific profanity filtering

lib/cdo/profanity_filter.rb

@@ -0,0 +1,33 @@
+require 'cdo/web_purify'
+
+class ProfanityFilter
+  # List of words that should be allowed only in specific languages.
+  #
+  # Entries are in the format
+  #   foobar: %w(en es)
+  # Where the key "foobar" is the word in question,
+  # and the value is a list of languages that should allow that word.
+  #
+  # Words in this list will be blocked for all languages _except_ for those
+  # listed along with the word.
+  # Words in this list should be _unblocked_ on WebPurify, since we are
+  # handling them with our custom code, here.
+  LANGUAGE_SPECIFIC_ALLOWLIST = {
+    fu: %w(it), # past-tense "to be" in Italian
+    fick: %w(sv) # "got" in Swedish
+  }
+
+  # Look for profanity in a given text, return the first expletive found
+  # or nil if no profanity is found.
+  #
+  # @param [String] text to check for profanity
+  # @param [String] language_code a two-character ISO 639-1 language code
+  def self.find_potential_profanity(text, language_code)
+    LANGUAGE_SPECIFIC_ALLOWLIST.each do |word, languages|
+      next if languages.include? language_code
+      r = Regexp.new "\\b#{word}\\b", Regexp::IGNORECASE
+      return word.to_s if r =~ text
+    end
+    WebPurify.find_potential_profanity(text, ['en', language_code])
+  end
+end

lib/cdo/share_filtering.rb

@@ -1,6 +1,6 @@
 require 'cdo/regexp'
 require 'cdo/geocoder'
-require 'cdo/web_purify'
+require 'cdo/profanity_filter'
 require 'dynamic_config/gatekeeper'
 
 USER_ENTERED_TEXT_INDICATORS = ['TITLE', 'TEXT', 'title name\=\"VAL\"'].freeze
@@ -23,6 +23,9 @@ module FailureType
   #
   # May throw OpenURI::HTTPError, IO::EAGAINWaitReadable depending on
   # service availability.
+  #
+  # @param [String] program the student's program text
+  # @param [String] locale a two-character ISO 639-1 language code
   def self.find_share_failure(program, locale)
     return nil unless should_filter_program(program)
 
@@ -38,7 +41,7 @@ def self.find_share_failure(program, locale)
     phone_number = RegexpUtils.find_potential_phone_number(program_tags_removed)
     return ShareFailure.new(FailureType::PHONE, phone_number) if phone_number
 
-    expletive = WebPurify.find_potential_profanity(program_tags_removed, ['en', locale])
+    expletive = ProfanityFilter.find_potential_profanity(program_tags_removed, locale)
     return ShareFailure.new(FailureType::PROFANITY, expletive) if expletive
 
     nil

lib/test/cdo/test_share_filtering.rb

@@ -66,6 +66,72 @@ def test_find_share_failure_with_profanity
     )
   end
 
+  def test_profanity_with_italian_edge_case
+    # "fu" is a past-tense "to be" in Italian, but should be blocked
+    # as profanity in English.  WebPurify doesn't support this, so we
+    # have custom filtering that takes locale into account for this word.
+    program = generate_program('My Custom Profanity', 'fu')
+    innocent_program = generate_program('My Innocent Program', 'funny tofu')
+
+    # Stub WebPurify because we expect our custom blocking to handle this case.
+    WebPurify.stubs(:find_potential_profanity).returns(nil)
+
+    # Blocked in English
+    assert_equal(
+      ShareFailure.new(ShareFiltering::FailureType::PROFANITY, 'fu'),
+      ShareFiltering.find_share_failure(program, 'en')
+    )
+
+    # But the innocent program is fine
+    assert_nil(
+      ShareFiltering.find_share_failure(innocent_program, 'en')
+    )
+
+    # Blocked in Spanish
+    assert_equal(
+      ShareFailure.new(ShareFiltering::FailureType::PROFANITY, 'fu'),
+      ShareFiltering.find_share_failure(program, 'es')
+    )
+
+    # Allowed in Italian
+    assert_nil(
+      ShareFiltering.find_share_failure(program, 'it')
+    )
+  end
+
+  def test_profanity_with_swedish_edge_case
+    # "fick" means "got" in Swedish, but should be blocked
+    # as profanity in English.  WebPurify doesn't support this, so we
+    # have custom filtering that takes locale into account for this word.
+    questionable_program = generate_program('My Custom Profanity', 'fick')
+    innocent_program = generate_program('My Innocent Program', 'fickle')
+
+    # Stub WebPurify because we expect our custom blocking to handle this case.
+    WebPurify.stubs(:find_potential_profanity).returns(nil)
+
+    # Blocked in English
+    assert_equal(
+      ShareFailure.new(ShareFiltering::FailureType::PROFANITY, 'fick'),
+      ShareFiltering.find_share_failure(questionable_program, 'en')
+    )
+
+    # But the innocent program is fine
+    assert_nil(
+      ShareFiltering.find_share_failure(innocent_program, 'en')
+    )
+
+    # Blocked in Italian
+    assert_equal(
+      ShareFailure.new(ShareFiltering::FailureType::PROFANITY, 'fick'),
+      ShareFiltering.find_share_failure(questionable_program, 'it')
+    )
+
+    # Allowed in Swedish
+    assert_nil(
+      ShareFiltering.find_share_failure(questionable_program, 'sv')
+    )
+  end
+
   def test_find_share_failure_for_non_playlab
     program = '<xml><block type=\"controls_repeat\">'\
       '<title name=\"TIMES\">4</title><statement name=\"DO\">'\