tommy/libpostal
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

[fix] genitive case for Russian/Ukrainian toponyms, not locative (#125)

Browse Source
This commit is contained in:
Al
2016-12-28 14:34:20 -05:00
parent e6e4b28e43
commit 7ee44a584b
2 changed files with 12 additions and 12 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
resources/parser/default.yaml
View File

@@ -121,12 +121,12 @@ state:
full_name_probability: 0.2 full_name_probability: 0.2
abbreviated_probability: 0.8 abbreviated_probability: 0.8
# Currently for Russian and Ukrainian, convert some names to the genitive/locative case # Currently for Russian and Ukrainian, convert some names to the genitive case
slavic_names: slavic_names:
state: state:
locative_probability: 0.4 genitive_probability: 0.4
state_district: state_district:
locative_probability: 0.4 genitive_probability: 0.4
country: country:
# If no country is specified, pull the country name from CLDR (authoratative country names translated into different languages) # If no country is specified, pull the country name from CLDR (authoratative country names translated into different languages)

18
scripts/geodata/addresses/components.py
View File

@@ -858,31 +858,31 @@ class AddressComponents(object):
else: else:
return self.japanese_node_admin_level_map.get(val.get('place'), 1000) return self.japanese_node_admin_level_map.get(val.get('place'), 1000)
def locative_name(self, name, language): def genitive_name(self, name, language):
morph = self.slavic_morphology_analyzers.get(language) morph = self.slavic_morphology_analyzers.get(language)
if not morph: if not morph:
return None return None
norm = [] norm = []
words = safe_decode(name).split() words = safe_decode(name).split()
n = len(words) n = len(words)
for i, word in enumerate(words):
for word in words:
parsed = morph.parse(word)[0] parsed = morph.parse(word)[0]
word_class = {'gent'} if i < n - 1 else {'loct'} inflected = parsed.inflect({'gent'})
inflected = parsed.inflect(word_class)
if inflected and inflected.word: if inflected and inflected.word:
norm.append(inflected.word) norm.append(inflected.word)
else: else:
norm.append(word) norm.append(word)
return six.u(' ').join(norm) return six.u(' ').join(norm)
def add_locatives(self, address_components, language): def add_genitives(self, address_components, language):
if language in self.slavic_morphology_analyzers and AddressFormatter.CITY in address_components: if language in self.slavic_morphology_analyzers and AddressFormatter.CITY in address_components:
for component in address_components: for component in address_components:
if component not in AddressFormatter.BOUNDARY_COMPONENTS: if component not in AddressFormatter.BOUNDARY_COMPONENTS:
continue continue
locative_probability = nested_get(self.config, ('slavic_names', component, 'locative_probability'), default=None) genitive_probability = nested_get(self.config, ('slavic_names', component, 'genitive_probability'), default=None)
if locative_probability is not None and random.random() < float(locative_probability): if genitive_probability is not None and random.random() < float(genitive_probability):
address_components[component] = self.locative_name(address_components[component], language) address_components[component] = self.genitive_name(address_components[component], language)
def abbreviated_state(self, state, country, language): def abbreviated_state(self, state, country, language):
abbreviate_state_prob = float(nested_get(self.config, ('state', 'abbreviated_probability'))) abbreviate_state_prob = float(nested_get(self.config, ('state', 'abbreviated_probability')))
@@ -1719,7 +1719,7 @@ class AddressComponents(object):
self.drop_invalid_components(address_components, country) self.drop_invalid_components(address_components, country)
self.add_locatives(address_components, language) self.add_genitives(address_components, language)
if language_suffix and not non_local_language and not language_altered: if language_suffix and not non_local_language and not language_altered:
language = language_suffix.lstrip(':').lower() language = language_suffix.lstrip(':').lower()