Al
22c4e99ea0
[parser] As part of reading/tokenizing the address parser data set,
...
several copies of the same training example will be generated.
1. with only lowercasing
2. with simple Latin-ASCII normalization (no umlauts, only things that
are common to all languages)
3. basic UTF-8 normalizations (accent stripping)
4. language-specific Latin-ASCII transliteration (e.g. ü => ue in German)
This will apply both on the initial passes when building the phrase
gazetteers and during each iteration of training. In this way, only the
most basic normalizations like lowercasing need to be done at runtime
and it's possible to use only minimal normalizations like lowercasing.
May have a small effect on randomization as examples are created in a
deterministic order. However, this should not lead to cycles since the
base examples are shuffled, thus still satisfying the random permutation
requirement of an online/stochastic learning algorithm.
2016-12-02 13:09:03 -05:00
Al
4b35da629f
[numex] regenerated numex data file
2016-11-30 15:58:55 -08:00
Al
4677874610
[parser] stripping postal codes of phrases like CP (in Spanish) before adding them to the gazetteers, whether it's concatenated or a separate token. Adding a command-line argument for the number of iterations
2016-11-30 15:58:03 -08:00
Al
0e29cdd9fd
[parser] fixing some uninitialized value issues during parser training
2016-11-30 15:42:09 -08:00
Al
f5a6bd0f36
[fix] sparse_matrix_new_from_matrix uses new matrix types
2016-11-30 10:15:12 -08:00
Al
b639fa5127
[utils] string_replace also creates a copy
2016-11-30 10:09:33 -08:00
Al
89f6611c4e
[strings] string_trim makes a copy rather than modifying the pointer
2016-11-28 15:06:07 -08:00
Al
d922d9a60a
[expansion] regenerated address_expansion_data.c
2016-11-28 10:47:15 -08:00
Al
f78281456a
[fix] header defintion
2016-11-27 01:00:25 -08:00
Al
eea11beb6a
[expansion] using easier-to-access data structure for address dictionaries
2016-11-27 00:56:48 -08:00
Al
7298c895c8
[utils] adding a chunked shuffle as the concatenated file sizes may get larger than memory
2016-11-21 14:04:34 -05:00
Al
58851a9088
[normalization] Adding NORMALIZE_STRING_SIMPLE_LATIN_ASCII option so parser can normalize punctuation and HTML entities, etc. without touching the alphanumeric parts of the original input
2016-08-21 19:45:32 -04:00
Al
8b9702b43d
[error handling] Checking that resize succeeded in transliterate.c
2016-08-21 19:43:09 -04:00
Al
2644fed18f
[transliteration] Adding LATIN_ASCII_SIMPLE constant to transliterate.h
2016-08-21 19:42:10 -04:00
Al
4375bdea3b
[transliteration] strduping transliterator name while building table
2016-08-21 19:41:34 -04:00
Al
bde8776bc2
[transliteration] Regenerating transliteration data files
2016-08-21 19:41:11 -04:00
Al
330edc2c93
[utils] cstring_array_get_phrase requires a char_array to be passed in so it doesn't have to do any memory allocation
2016-08-16 13:11:45 -04:00
Al
92e66fd60c
[utils] string_next_hyphen_index
2016-08-16 12:49:52 -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
20aad99a38
[parser] enum just lists boundary types
2016-07-30 17:07:23 -04:00
Al
965bac1833
[trie] Making methods to construct string phrases from phrase matches available through trie_search.h
2016-07-30 17:06:20 -04:00
Al
08f39d6b80
[parser] Adding address_parser_rewind to make multiple passes through the file when compiling the phrase tries
2016-07-28 17:13:58 -04:00
Al
1b09b7f2e5
[fix] Adding country_region to address_parser_train
2016-07-28 16:18:32 -04:00
Al
c6af5cc071
[parser] Adding country_region label to parser as a boundary component
2016-07-28 15:19:48 -04:00
Al
64f167f045
[tokenization] Re-generating scanner
2016-07-21 17:04:57 -04:00
Al
81b4a4a1cb
[tokenization] Hyphens, etc. between non-ASCII digits (e.g. Unicode full-width numbers) should be single tokens
2016-07-21 17:04:57 -04:00
Al
be5fd79a48
[expansion] Prefix/suffix expansions by default can apply to ADDRESS_ANY but also inherit the types of any dictionary that lists their canonical form (so we can add suffixes without worrying about whether they're for streets or place names, etc.)
2016-07-21 17:04:57 -04:00
Al
8926293063
[parser/cli] Using NFC normalization on the output in the parser client ( closes #30 ). Optional command-line arg for parser output dir, useful for spot-checking different experiments
2016-07-21 17:04:57 -04:00
Al
44908ff95a
[parser] No digit normalization in training data-derived parser phrases (for postcodes, etc.), phrases include the new island type, house number phrases if any are valid. Adjacent words are now full phrases if they are part of a multiword token like a city name. For hyphenated names like Carmel-by-the-Sea, adding a version to the phrase dictionary where the hyphens are replaced with spaces
2016-07-21 17:04:57 -04:00
Al
41ae742285
[fix] tokenized trie search when falling off the trie at the start of a valid phrase
2016-07-21 17:04:57 -04:00
Al
6e60b3bbda
[fix] semicolon in #define
2016-07-21 17:04:57 -04:00
Al
b5d4dd6f37
[tokenization] Including full-width numbers in numeric tokens
2016-07-21 17:04:57 -04:00
Al
dd7ef6fabf
[dictionaries] Making new component for near/nearby prepositions
2016-07-21 17:04:57 -04:00
Al
2454b98c6d
[tokenization] Reverting commit for tokenizing initial/final apostrophes as part of words as it may be more effective to handle during post-processing
2016-07-21 17:04:57 -04:00
Al
0a8f46bdc3
[parser] Using new geonames designations in parser features
2016-07-21 17:04:57 -04:00
Al
c383f8af88
[parser] Using NFC normalization for parser as well, @ sign not defined as separator since it may also be used in intersections
2016-07-21 17:04:57 -04:00
Al
c2ee5a45b3
[geodb] Adding separate bitset for geonames place types and using NFC normalization instead of NFD (requires retraining)
2016-07-21 17:04:57 -04:00
Al
6c39c663ff
[normalize] Adding NORMALIZE_STRING_COMPOSE for NFC unicode normalization
2016-07-21 17:04:57 -04:00
Al
757c6147cb
[tokenization] Adding ability to tokenize 's Gravenhage
2016-07-21 17:04:57 -04:00
Al
2e8888e331
[fix] warnings/size_t in libpostal.c
2016-07-21 17:04:57 -04:00
Al
e800f21f06
[gazetteers] Adding new gazetteer types/address components
2016-07-21 17:04:57 -04:00
Al
e5e0cf3b92
[fix] loading transliteration module in address_parser_test.c as well
2016-07-21 17:04:57 -04:00
Al
b8d43dc601
[fix] cstring_array_split calls
2016-07-21 17:04:57 -04:00
Al
b19cd3f60a
[fix] brace
2016-07-21 17:04:57 -04:00
Al
994b2f18e4
[parser] Ignore multiple spaces in parser input post-normalization. If normalizing the string creates several distinct tokens (namely in Vulgar fractions e.g. ½ => 1/2), add all the sub-tokens with the same label as the parent
2016-07-21 17:04:57 -04:00
Al
b664ab1cea
[utils] Adding cstring_array_split_ignore_consecutive
2016-07-21 17:04:57 -04:00
