[languages] Use English formats for Romanized CJK

resources/parser/default.yaml

@@ -7,6 +7,11 @@ languages:
     # Replace user-tagged admin components with the non-local language version
     replace_non_local_probability: 0.4
 
+    language_code_replacements:
+        ja_rm: en
+        ko_rm: en
+        zh_pinyin: en
+
 # Dependencies for including each component in an "address"
 # Two-way dependencies are not an issue
 component_dependencies:

scripts/geodata/addresses/components.py

@@ -136,6 +136,8 @@ class AddressComponents(object):
         # Non-admin component dropout
         self.address_level_dropout_probabilities = {k: v['probability'] for k, v in six.iteritems(self.config['dropout'])}
 
+        self.language_code_replacements = nested_get(self.config, ('languages', 'language_code_replacements'))
+
         self.osm_admin_rtree = osm_admin_rtree
         self.language_rtree = language_rtree
         self.neighborhoods_rtree = neighborhoods_rtree
@@ -1105,6 +1107,10 @@ class AddressComponents(object):
 
         self.drop_invalid_components(address_components)
 
+        if language_suffix:
+            suffix_lang = language_suffix.lstrip(':')
+            language = self.language_code_replacements.get(suffix_lang, language)
+
         return address_components, country, language
 
     def limited(self, address_components, latitude, longitude):