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

[fix] component dependencies

Browse Source
This commit is contained in:
Al
2016-05-21 19:06:39 -04:00
parent 4e9906fcab
commit a4c215ce9a
1 changed files with 4 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
scripts/geodata/addresses/components.py
View File

@@ -138,7 +138,7 @@ class AddressComponents(object):
forward_deps = self.config.get('component_dependencies', {}) forward_deps = self.config.get('component_dependencies', {})
for i, component in enumerate(forward_deps): for i, component in enumerate(AddressFormatter.address_formatter_fields):
self.component_bit_values[component] = 1 << i self.component_bit_values[component] = 1 << i
all_values = self.component_bitset(forward_deps) all_values = self.component_bitset(forward_deps)
@@ -169,6 +169,9 @@ class AddressComponents(object):
be invalid to have an address that was simply a house number with no other information. The be invalid to have an address that was simply a house number with no other information. The
caller of this method may decide to drop all the components at once or one at a time, creating caller of this method may decide to drop all the components at once or one at a time, creating
N training examples from a single address. N training examples from a single address.
Some components are also more likely to be dropped than others, so in the same config there are
dropout probabilities for each.
''' '''
component_bitset = self.component_bitset(components) component_bitset = self.component_bitset(components)