Al
|
8a5da5f860
|
[boundaries/osm] Reverting admin_level=10 back to city_district for India so it'll match the current training data, can revisit later
|
2016-08-13 22:51:42 -04:00 |
|
Al
|
bc8acb196c
|
[osm] Pulling valid postal codes out into a method
|
2016-08-13 01:49:26 -04:00 |
|
Al
|
55895369b8
|
[boundaries] Using state again for UK countries (England, Scotland, Wales, Northern Ireland). country_region was created mostly for non-administrative regions of a country (usually admin_level=3 in OSM). The UK is a bit more complicated in that there are multiple non-sovereign countries, but it's probably not worth creating a different tag and different set of parameters just to have a distinct name for 1st level admin in the UK
|
2016-08-11 23:47:31 -04:00 |
|
Al
|
d51a6693ac
|
[fix] reverting commit that was lumped in with geonames script
|
2016-08-11 21:49:29 -04:00 |
|
Al
|
74d042e3c7
|
[boundaries] For India, making admin_level 10 map to suburb rather than city_district
|
2016-08-11 21:47:10 -04:00 |
|
Al
|
29081a0699
|
[fix] adding English template insertions for the UK regardless of language
|
2016-08-11 21:32:54 -04:00 |
|
Al
|
22123b80ba
|
[fix] refactoring geonames script a bit
|
2016-08-11 21:31:39 -04:00 |
|
Al
|
48755ec218
|
[boundaries] Adding regex replacements for boundary names such as Lyon 2e Arrondissement where putting Lyon is the OSM convention but we might sometimes want just 2e Arrondissement to appear in the training data next to Lyon
|
2016-08-11 13:09:24 -04:00 |
|
Al
|
10a41309b8
|
[addresses] Increasing Romaji probability to 0.4
|
2016-08-06 21:27:32 -04:00 |
|
Al
|
b993e9a163
|
[fix] add Japanese-language variant if metro station is added
|
2016-08-06 21:17:14 -04:00 |
|
Al
|
39bd562d04
|
[addresses] only set language if we needed it for Japanese house_numbers
|
2016-08-06 21:06:01 -04:00 |
|
Al
|
cdd5a96346
|
[addresses] metro station can also be used for plain venues without a house number so we get more in the training set
|
2016-08-06 20:52:29 -04:00 |
|
Al
|
5ec752e887
|
[fix] order of ops
|
2016-08-06 20:43:13 -04:00 |
|
Al
|
e68fee7c68
|
[fix] null check
|
2016-08-06 20:39:28 -04:00 |
|
Al
|
3e34012e69
|
[fix] if the language is given already, use it as a suffix rather than choosing at random
|
2016-08-06 20:36:56 -04:00 |
|
Al
|
606c464db6
|
[fix] house number phrases
|
2016-08-06 20:11:32 -04:00 |
|
Al
|
e35649f09d
|
[fix] import
|
2016-08-06 20:01:38 -04:00 |
|
Al
|
0e7cb2b06c
|
[fix] var name II
|
2016-08-06 20:00:35 -04:00 |
|
Al
|
8d88820d30
|
[fix] var name
|
2016-08-06 19:59:53 -04:00 |
|
Al
|
374c46ada5
|
[fix] metro station properties
|
2016-08-06 19:56:13 -04:00 |
|
Al
|
0edfbe0d61
|
[osm] Adding metro stations index to training data options
|
2016-08-06 19:52:21 -04:00 |
|
Al
|
195278cfea
|
[osm] Reverse geocoding to metro station only for addresess in Japan
|
2016-08-06 19:50:18 -04:00 |
|
Al
|
6ef54bcc6f
|
[addresses] Adding metro stations to AddressComponents expansion
|
2016-08-06 19:36:57 -04:00 |
|
Al
|
da2985a4ae
|
[places] Metro station dropout probabilities
|
2016-08-06 19:34:56 -04:00 |
|
Al
|
6ce882cb55
|
[addresses] Metro station component dependencies (road or house_number)
|
2016-08-06 19:34:39 -04:00 |
|
Al
|
668aa20996
|
[addresses] Metro station phrases for Japanese Romaji
|
2016-08-06 19:34:07 -04:00 |
|
Al
|
9cbbca5e47
|
[addresses] Metro station phrase for Japanese
|
2016-08-06 19:33:42 -04:00 |
|
Al
|
d59ab82701
|
[metro stations] Adding metro station phrase generator
|
2016-08-06 19:33:21 -04:00 |
|
Al
|
1e27ad1124
|
[metro stations] Adding metro station component to address formatter
|
2016-08-06 19:13:20 -04:00 |
|
Al
|
5cff119d25
|
[fix] command line arg
|
2016-08-06 18:36:27 -04:00 |
|
Al
|
406666362c
|
[fix] command-line index creation
|
2016-08-06 18:36:01 -04:00 |
|
Al
|
7ddd553129
|
[fix] metro stations reverse geocoder
|
2016-08-06 18:30:54 -04:00 |
|
Al
|
5e44f6954b
|
[metro stations] Adding metro stations reverse geocoder
|
2016-08-06 18:24:25 -04:00 |
|
Al
|
954bb08a8d
|
[points] Fixes to point index
|
2016-08-06 18:23:30 -04:00 |
|
Al
|
964728a02d
|
[fix] block phrases for Japanese and namespaced language handling in case Romaji is chosen before normalization
|
2016-08-06 14:50:39 -04:00 |
|
Al
|
684550ea7d
|
[fix] only add house_number phrase to numeric inputs
|
2016-08-06 14:49:28 -04:00 |
|
Al
|
8b5d44e173
|
[fix] Japanese house numbers aren't without dependencies, just have different ones (road or suburb or city_district)
|
2016-08-06 03:38:44 -04:00 |
|
Al
|
2c024ce9f4
|
[addresses] special case for Japan, house_number does not depend on street name
|
2016-08-06 02:38:58 -04:00 |
|
Al
|
445e8082c8
|
[addresses] Adding per-country overrides for address component dependencies
|
2016-08-06 02:36:47 -04:00 |
|
Al
|
3137ef5c6a
|
[build] configure/Makefile changes to use SIMD exp and BLAS when available
|
2016-08-06 00:43:24 -04:00 |
|
Al
|
59e28c6c2a
|
[math] double_array definition in collections.h to use new vectorized exp
|
2016-08-06 00:40:38 -04:00 |
|
Al
|
46cd725c13
|
[math] Generic dense matrix implementation using BLAS calls for matrix-matrix multiplication if available
|
2016-08-06 00:40:01 -04:00 |
|
Al
|
d4a792f33c
|
[math] Adding fast SIMD exponent using the Remez algorithm for vectorized exp
|
2016-08-06 00:31:16 -04:00 |
|
Al
|
161f18575d
|
[utils] Adding realloc checks to vector implementation
|
2016-08-05 23:02:52 -04:00 |
|
Al
|
14c35b35c6
|
[fix] probabilities in Romanian address config
|
2016-08-04 17:53:10 -04:00 |
|
Al
|
13718355cc
|
[test] Test zones in address configs
|
2016-08-04 17:52:19 -04:00 |
|
Al
|
eb4c957b4c
|
[test] Adding tests for known number of floors as it touches different parts of the address configs
|
2016-08-03 17:40:48 -04:00 |
|
Al
|
f33882b7bc
|
[fix] Swedish config for top floor phrase
|
2016-08-03 11:54:15 -04:00 |
|
Al
|
813f29f299
|
[osm] Removing the call to normalize_place_names in place data formatting as we should be able to trust the places more than the addresses
|
2016-08-02 16:29:34 -04:00 |
|
Al
|
0ab3b13b75
|
[osm] Remove hanging commas, slashes, etc. Implementing a stricter rule for user-specified tags (not reverse geocoded) so that if they contain an unknown phrase followed by an unknown boundary phrase, we delete that tag and fall back to the reverse geocoded components. Moving CLDR country tagging to later in the process since those are known correct names.
|
2016-08-02 16:25:45 -04:00 |
|