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

[parser] initial version of new parser address config for English

Browse Source
This commit is contained in:
Al
2016-03-28 02:23:32 -04:00
parent c183db1020
commit 161012f9f4
1 changed files with 343 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

343
resources/addresses/en.yaml Normal file
View File

@@ -0,0 +1,343 @@
# en.yaml
# -------
# Supplement to the per-country address formats for English around the world.
# Note: by default, we use the UK conventions as they cover more countries
# US/Canada-specific conventions and any others (e.g. Hong Kong, Australia)
# go in country overrides
#
# Remember, these rules only get applied sometimes with random probabilities
# unless specified using "frequency: always" (which should almost never be used)
# Number
# ======
# Number, No., #, etc. can be used in both floor and apartment numbers,
# so we'll define it separately
number:
canonical: number # canonical word in libpostal dictionary
abbreviated: "no" # abbreviated form. no is a boolean in YAML, needs to be quoted
prefer_abbreviated: true # Use abbreviated form more often than the canonical form
sample: true # Randomly sample other variations (e.g. num, nr)
numeric_abbreviated:
affix: "#" # e.g. #3, #2F, etc.
direction: left # affix goes on the number's left
# Floor/level
# ===========
# OSM doesn't usually concern itself with the address beyond the front door
# yet many real-world addresses will have qualifying strings like "6th floor"
# and we'd like the parser to handle those.
#
# When we do get floor numbers in OSM addresses, it's usually in the form of the
# addr:floor or level tag, where the value is typically an integer or a half-floor
# (to indicate mezzanines). Those tags are relatively scarce in OSM, but many OSM
# addresses do have a building:levels tag. If we know there are 20 floors in the
# building, we can randomly sample numbers <= the # of floors and come up with plausible
# sounding addresses (i.e. a Floor 20 address is not as likely outside major cities).
#
# We're not done yet, because the integer value by itself isn't what people use when
# writing addresses. This part of the config helps us rewrite the raw integer floor
# numers as the sort of natural language text used in addresses like "Fl #1". The config
# is designed to be cross-lingual, so we can use the same structure with different words
# and do this for addresses in pretty much any language.
level:
# Numbered floors
floor:
canonical: floor
abbreviated: fl
sample: true
# e.g. Floor 1
numeric:
direction: left # Floor/Fl goes to the left of the number
use_number_phrase: true # Occasionally add variation of "number", e.g. Floor #1, Floor No. 1
# e.g. 2/F, 3/F
numeric_abbreviated:
affix: /f
direction: right # affix goes to number's right
# e.g. 1st Floor
ordinal:
direction: right
# Special instructions for ground floor
ground_floor:
number: 0 # the 0th floor is typically the ground level in the UK/Commonwealth
canonical: ground floor
abbreviated: g/f
sample: true
# Special instructions for lower ground floor (added randomly, not an alias for a floor number)
lower_ground_floor:
canonical: lower ground floor
abbreviated: lg
sample: true
# Special instructions for upper ground floor (added randomly, not an alias for a floor number)
upper_ground_floor:
canonical: upper ground floor
abbreviated: ug
sample: true
# Special instructions for podium level (added randomly)
podium_level:
canonical: podium level
abbreviated: pd lvl
sample: true
alternates:
- canonical: podium
abbreviated: pd
sample: true
# Used when floor number is < 0 (starts at -1 in all countries)
basement:
canonical: basement
abbreviated: bsmt
sample: true
# e.g. Basement 1
numeric:
direction: left
# e.g. B1
numeric_abbreviated:
affix: b
direction: left
# e.g. 2nd Basement
ordinal:
direction: right
# Floor number of <= -2 can be basement 2, sub-basement, sub-basement 1, etc.
sub_basement:
canonical: sub basement
abbreviated: sb
sample: true
# e.g. Sub-basement 1
numeric:
direction: left
# e.g. SB1
numeric_abbreviated:
affix: sb
direction: left
# e.g. 2nd Sub-basement
ordinal:
direction: right
# Mezzanine level (floor number {0.5, 1.5, ...}, also be added at random)
mezzanine:
# Floor 0.5 is just plain mezzanine, no number
canonical: mezzanine
abbreviated: mezz
half_floors: true
sample: true
# Mezzanine/Mezz 2
numeric:
direction: left
# M2
numeric_abbreviated:
affix: m
direction: left
# 2nd Mezzanine
ordinal:
direction: right
# Intersections
# =============
# For constructing intersections like 5th Avenue & Broadway
# In OSM, a node that's part of two ways is an intersection.
#
# These simple rules make it possible to create training examples
# like: 26th/road Street/road and/intersection 6th/road Avenue/road
intersections:
# 26th & 6th Avenue
and:
canonical: and
abbreviated: "&"
sample: true
# 26th @ Broadway
at:
canonical: at
abbreviated: "@"
sample: true
# 26th betw 5th Ave and 6th Ave
between:
canonical: between
abbreviated: betw
sample: true
# PO Box addresses
# ================
# For PO box addresses, there's almost no data in OSM, so we'll need to
# generate them somewhat randomly.
#
# The strategy is: for every amenity=post_office, generate a number of PO box
# addresses using random numbers (and some alpha-numerics so we capture patterns
# like PO Box 1Q, etc.) It doesn't matter if the post boxes themselves actually
# exist, as long as they cover the patterns of digits we expect in real addresses.
# The parser cares more about how many digits a number has and the surrounding
# words/phrases than the specific number i.e. numbers in the range 1000-9999
# can simply be normalized to DDDD.
po_box:
numeric:
# Don't sample all the forms in post_office.txt as many of the PO box
# phrases appear only in Australia
sample: false
default:
canonical: post office box
abbreviated: po box
sample: true
prefer_abbreviated: true
# Alternative phrases to use
alternates:
- canonical: box
sample: true
- canonical: private mail box
abbreviated: pmb
prefer_abbreviated: true
sample: true
# Directions
# ==========
# Unit types, stairways, etc. may have a direction associated
# with them whether it's right/left or a cardinal direction
# like "East Entrance".
directions:
right:
canonical: right
numeric_abbreviated:
affix: r
direction: right
left:
canonical: left
numeric_abbreviated:
affix: l
direction: right
rear:
canonical: rear
numeric_abbreviated:
affix: r
direction: right
front:
canonical: front
numeric_abbreviated:
affix: f
direction: right
east:
canonical: east
numeric_abbreviated:
affix: e
direction: right
west:
canonical: west
numeric_abbreviated:
affix: w
direction: right
north:
canonical: north
numeric_abbreviated:
affix: n
direction: right
south:
canonical: south
numeric_abbreviated:
affix: s
direction: right
# Unit types
# ==========
# Unit information is common in residential addresses, offices, business parks, etc.
# Just like thoroughfare types (Street, Avenue, etc.), there are many common ways to
# refer to the
unit_types:
# Units are not part of the global address formats (and are not always standard)
# This is a list of places in the address where the unit line might go
order:
# e.g. Flat 18, Da Vinci House, 44 Saffron Hill, London
- before: house
# e.g. Da Vinci House, Flat 18, 44 Saffron Hill, London
- before: road
# e.g. 123 East 45th St, Apt 6, NYC
- after: road
# Special terms
suite: &suite
canonical: suite
abbreviated: ste
sample: true
penthouse: &penthouse
canonical: penthouse
abbreviated: ph
sample: true
office: &office
canonical: office
abbreviated: ofc
sample: true
door: &door
canonical: door
sample: true
room: &room
canonical: room
abbreviated: rm
sample: true
apartment: &apartment
canonical: apartment
abbreviated: apt
prefer_abbreviated: true
sample: true
flat: &flat
canonical: flat
abbreviated: fl
sample: true
lot: &lot
canonical: lot
sample: true
unit: &unit
canonical: unit
abbreviated: u
use_number_phrase: true
sample: true
numeric:
# Many unit types that apply only in Australia
# For most English-speaking countries, only use the terms defined above
sample: false
use_number_phrase: true
use_direction_phrase: true
default: *flat
alternates:
- *apartment
- *unit
- *door
- *room
- *office
- *penthouse
- *lot
# Country-specific overrides
# ==========================
# For each country, we allow a copy of the structures listed above
# in order to override the default values
countries:
# United States
us:
level:
ground_floor:
number: 1
unit_types:
numeric:
default: *apartment
country_alternates:
- *flat
# Canada
# Note: this is Canadian English only. If the address is in French it will use the French config
ca:
level:
ground_floor:
number: 1
unit_types:
numeric:
default: *apartment
country_alternates:
- *flat
# Australia
au:
po_box:
numeric:
# Australia has many strings for this e.g. Roadside Mail Bag
sample: true
unit_types:
numeric:
# Australia has all kinds of unit types (e.g. Marine Berth) not used elsewhere
sample: true