Al
|
b90c3dab4b
|
[similarity/dedupe] adding Soft-TFIDF implementation with several different fallback qualifiers for the max-sim function (Damerau-Levenshtein and libpostal's new bucketed affine gap method for detecting abbreviations), but keeping Jaro-Winkler as the secondary similarity function in the final distance metric. Overall this should results in higher similarity values when one of the tokens may not quite match the pure secondary threshold in terms of Jaro-Winkler but may match on one of the other criteria.
|
2017-12-28 04:34:46 -05:00 |
|
Al
|
33bb90d94b
|
[utils] adding place.h header, which converts parser-like output into an object that can be used for comparisons. Currently single-value, but could use cstring_arrays for fields instead.
|
2017-12-27 22:13:04 -05:00 |
|
Al
|
d731339811
|
[expand] fixing case where too many permutations were getting added for longer strings due to the new-ish ordinal suffix handling, using string_tree_num_tokens instead of string_tree_num_strings throughout to check for previously added words, using new is_likely_roman_numeral API
|
2017-12-27 21:48:54 -05:00 |
|
Al
|
b4fdc51bf9
|
[numex] changing is_roman_numeral to is_likely_roman_numeral to get rid of most of the false positives like \"La\" in Spanish which could be L(=50) + the ordinal suffix \"a\", but in practice it never means that. For Roman numerals that are shorter than two characters (whether on their own like "DC" or "MD", or attached to a potential ordinal suffix like \"Ce\" in French), will be ignored unless they're composed of more likely, smaller, Roman numerals: I, V, and X, so VI, IX, etc. are expanded as Roman numerals but LI is not.
|
2017-12-27 19:38:02 -05:00 |
|
Al
|
bb9f6a4c6b
|
Merge branch 'master' into lieu_api
|
2017-12-26 10:47:29 -05:00 |
|
Gregory Oschwald
|
1bb6278446
|
Fix leak of normalized value in early return
|
2017-12-25 19:33:07 -08:00 |
|
Al
|
152761fcbc
|
[expand] adding improvements to root expansions (using possible phrase roots even if they're abbreviated e.g. "E Ctr St", adding special valid components check for root expansions beyond what's stored in the build address dictionaries), removing spaces before checking unique strings, only splitting numeric from alpha in the case of non-ordinals, using cstring_array internally and char ** in the public API
|
2017-12-25 01:37:42 -05:00 |
|
Al
|
a3f39be0d4
|
[fix] reverting gazetteer changes as it would affect the parser features as well and require retraining
|
2017-12-24 23:51:35 -05:00 |
|
Al
|
acbebc9ecf
|
[build] adding new source files for near dupe hashing and the command-line program to the Makefile
|
2017-12-24 12:44:10 -05:00 |
|
Al
|
f3a626463a
|
[api] adding API functions for near dupe hashes to the public header
|
2017-12-24 12:43:28 -05:00 |
|
Al
|
8b75c44026
|
[dedupe] adding a test program for near dupe hashing that simply prints out the results. Automated tests in the works
|
2017-12-24 12:41:44 -05:00 |
|
Al
|
acfdb50d7c
|
[dedupe] adding near-dupe hashing function, which can be thought of as the blocking function in record linkage or as a form of locally sensitive hashing in general document deduping. The goal is, if two addresses/names are the same, they should share at least one hash. These hashes can also be used as an inverted index (DB, ES, hashtable, etc.). Uses the double metaphone for name words in Latin script (otherwise each individual token, and sequences of two tokens in the case of ideograms for e.g. Chinese, Japanese, Korean, etc.)
|
2017-12-24 02:47:45 -05:00 |
|
Al
|
6c6e5062e0
|
[gazetteers] removing stopwords, etc. from numeric type components, adding street type expansions to name components
|
2017-12-24 02:19:41 -05:00 |
|
Al
|
c78566c241
|
[utils] adding cstring_array_extend and string_tree_clear
|
2017-12-24 01:46:20 -05:00 |
|
Al
|
4e3d868bd0
|
[parser] adding label constants to address_parser header
|
2017-12-24 01:45:50 -05:00 |
|
Al
|
d03ce4e058
|
[expand] remove blank expansions and strip spaces
|
2017-12-18 18:17:16 -05:00 |
|
AeroXuk
|
f6157224ed
|
Removing console colors for Windows builds.
|
2017-12-18 18:33:50 +00:00 |
|
Al
|
f63a9cc579
|
[expand] adding number phrases as ignorable in PO boxes
|
2017-12-17 22:12:12 -05:00 |
|
Al
|
727469b736
|
[expand] no longer delete phrases in cases like "PH 1" for units, where there's a phrase that can accompany numbered units and thus be ignored similar to "Apt 1" but that phrase may also be a qualifier (i.e. Apt 1 and Penthouse 1 are not the same)
|
2017-12-17 21:57:25 -05:00 |
|
Al
|
a1db4d7734
|
[expand/normalize] the split_alpha_from_numeric option now applies to both e.g. A1 and 1A since we now strip out ordinal suffixes prior to normalization
|
2017-12-17 19:53:15 -05:00 |
|
Al
|
8b2a4d1ecf
|
[api] adding libpostal_expand_address_root to the public API. This will attempt to delete tokens that can be safely ignored. It's deterministic and rule-based, but is informed by libpostal's fairly comprehensive dictionaries, and should work relatively well across languages for deduping purposes.
|
2017-12-17 17:46:26 -05:00 |
|
Al
|
9eef46adee
|
[expand] in cases like "Avenue D" where there are two phrases, one is ambiguous (and canonical) but not necessarily edge-ignorable (pre/post-directional), allow deletion of the other token (so "Avenue" in this case). Also allows skipping in cases where the language classifier may predict a second language with some small probability, such as French for a short string like "Avenue D" (in addition to English). If the token was ignorable in the highest probability language, ignore it in both.
|
2017-12-17 17:24:27 -05:00 |
|
Al
|
3f7abd5b24
|
[expand] adding a method that allows hash/equality comparisons of addresses like "100 Main" with "100 S Main St." or units like "Apt 101" vs. "#101". Instead of expanding the phrase abbreviations, this version tries its best to delete all but the root words in a string for a specific component. It's probably not perfect, but does handle a number of edge cases related to pre/post directionals in English e.g. "E St" will have a root word of simply "E", "Avenue E" => "E", etc. Also handles a variety of cases where the phrase could be a thoroughfare type but is really a root word such as "Park Pl" or the famous "Avenue Rd". This can be used for near dupe hashing to catch possible dupes for later analysis. Note that it will normalize "St Marks Pl" and "St Marks Ave" to the same thing, which is sometimes warranted (if the user typed the wrong thoroughfare), but can also be reconciled at deduping time.
|
2017-12-17 15:48:11 -05:00 |
|
Al
|
d0364ab6fb
|
[expand] adding method for checking phrase is in multiple dictionaries, and a helper method for determining whether an address phrase has a canonical interpretation
|
2017-12-17 03:14:00 -05:00 |
|
Travis
|
e1d89b5d62
|
[auto][ci skip] Adding data files from Travis build #323
|
2017-12-12 14:57:29 +00:00 |
|
Al
|
8968a6c966
|
[expand] moving expand to its own module so the internal methods can be exposed, calling from libpostal.c
|
2017-12-08 16:26:13 -05:00 |
|
Al
|
e4e84f0147
|
[utils] adding unicode_common_prefix/unicode_common_suffix, string_hyphen_prefix_len and string_hyphen_suffix_len to string_utils
|
2017-12-08 14:28:30 -05:00 |
|
Al
|
55ba627c3c
|
[similarity] needed to add utf8proc_category and invert the indices for counting transposes in affine gap
|
2017-12-08 14:27:23 -05:00 |
|
Al
|
cfa5b1ce42
|
[similarity] adding a stopword-aware acronym alignment method for matching U.N. with United Nations, Museum of Modern Art with MoMA, as well as things like University of California - Los Angeles with UCLA. All of these should work across languages, including non-Latin character sets like Cyrllic (but not ideograms as the concept doesn't make as much sense there). Skipping tokens like "of" or "the" depends only on the stopwords dictionary being defined for a given language.
|
2017-12-04 15:21:44 -05:00 |
|
Al
|
e27f5f1d70
|
[api] adding LIBPOSTAL_EXPORT to some of the new public API functions in this branch
|
2017-11-29 18:10:10 -05:00 |
|
Al
|
ec4d683d1b
|
Merge branch 'master' into lieu_api
|
2017-11-29 15:49:52 -05:00 |
|
Al
|
9e837f7209
|
[fix] deleting comment, this is not a header-only implementation
|
2017-11-29 12:26:36 -05:00 |
|
Al
|
26e4ef08bc
|
[fix] conditional compilation for strndup and drand48 for Windows, using config.h
|
2017-11-29 12:21:13 -05:00 |
|
Al
|
1a64ad682b
|
[merge] merging in the Ohio expansion numex changes from master
|
2017-11-29 11:51:43 -05:00 |
|
Al Barrentine
|
18eb5ef9ee
|
Merge pull request #272 from AeroXuk/master
Windows support via AppVeyor
|
2017-11-28 21:35:46 -05:00 |
|
AeroXuk
|
19ae97d527
|
Adding include config.h to strndup.c so that the function is not compiled and doesn't cause errors when the system has its own implementation.
|
2017-11-27 23:40:46 +00:00 |
|
AeroXuk
|
9090811826
|
Modifed the libpostal API to add an extra function libpostal_parser_print_features to toggle debugging info. Updated address_parser app to use the new function.
|
2017-11-27 19:20:37 +00:00 |
|
AeroXuk
|
69e0d5d963
|
Updated linenoise to be MSys2/MinGW compatible. Updated address_parser app to use the defined libpostal api and not include internal components directly. Removed windows src Makefile as it is now the same as the standard one.
|
2017-11-27 01:42:25 +00:00 |
|
AeroXuk
|
26ac9ab5c2
|
Removing EXPORT statements from all source files and most header files, leaving only the exports for the main API in libpostal.h. Modified Makefiles so that all the test apps build without having extra functions exported from libpostal.
|
2017-11-25 04:35:28 +00:00 |
|
Travis
|
15b3758be8
|
[auto][ci skip] Adding data files from Travis build #284
|
2017-11-24 22:29:45 +00:00 |
|
Al
|
ef098fd2e7
|
[numex] implementing the numex concat_only_if_number left context, which helps in the case of e.g. Columbus, OH in #271
|
2017-11-24 15:42:50 -05:00 |
|
Al
|
c276cf1529
|
[numex] adding a new type of left context for numeric expressions called conat_only_if_number (for something like "oh" which can be "Columbus, OH" or something like "Twenty-One Oh One"
|
2017-11-24 15:36:53 -05:00 |
|
AeroXuk
|
f0246e7333
|
Fix bug in strndup fix for windows. Move all includes out of headers and into code for strndup.h and move it to be the last include.
|
2017-11-23 19:11:25 +00:00 |
|
AeroXuk
|
f07ab765cb
|
Adding the export marker to all functions used in tests.
|
2017-11-20 20:58:37 +00:00 |
|
AeroXuk
|
ad682b7592
|
Altered Makefile to include strndup.c on the other programs which require it. For the windows version of the Makefile, commented out address_parser lines as it has dependencies on includes we don't have.
|
2017-11-20 20:24:11 +00:00 |
|
AeroXuk
|
2d3b420d35
|
Merging changes from AeroXuk/libpostal_windows.
|
2017-11-19 12:44:38 +00:00 |
|
Travis
|
7d6e648fc3
|
[auto][ci skip] Adding data files from Travis build #271
|
2017-11-17 19:36:25 +00:00 |
|
Al
|
fbf88aee88
|
[similarity] adding possible abbreviation functions to header, making everything const char *
|
2017-11-12 04:48:26 -05:00 |
|
Al
|
b34e578366
|
[similarity] using new sequence alignment breakdown by operation to tell if any two words are an abbreviation. The loose variant requires that the alignment covers all characters in the shortest string, which matches things like Services vs. Svc, whereas the strict variant requires that either the shorter string is a prefix of the longer one (Inc and Incorporated) or that the two strings share both a prefix and a suffix (Dept and Department). Both variants require that the strings share at least the first letter in common.
|
2017-11-11 04:02:28 -05:00 |
|
Al
|
751873e56b
|
[similarity] a *NEW* sequence alignment algorithm which builds on Smith-Waterman-Gotoh with affine gap penalties. Like Smith-Waterman, it performs a local alignment, and like the cost-only version of Gotoh's improvement, it needs O(mn) time and O(m) space (where m is the length of the longer string). However, this version of the algorithm stores and returns a breakdown of the number and specific types of edits it makes (matches, mismatches, gap opens, gap extensions, and transpositions) rather than rolling them up into a single cost, and without needing to return/compute the full alignment as in Needleman-Wunsch or Hirschberg's variant
|
2017-11-11 03:07:39 -05:00 |
|