Reduce size of language data (#4141)

* Move Turkish lemmas to a json file

Rather than a large dict in Python source, the data is now a big json
file. This includes a method for loading the json file, falling back to
a compressed file, and an update to MANIFEST.in that excludes json in
the spacy/lang directory.

This focuses on Turkish specifically because it has the most language
data in core.

* Transition all lemmatizer.py files to json

This covers all lemmatizer.py files of a significant size (>500k or so).
Small files were left alone.

None of the affected files have logic, so this was pretty
straightforward.

One unusual thing is that the lemma data for Urdu doesn't seem to be
used anywhere. That may require further investigation.

* Move large lang data to json for fr/nb/nl/sv

These are the languages that use a lemmatizer directory (rather than a
single file) and are larger than English.

For most of these languages there were many language data files, in
which case only the large ones (>500k or so) were converted to json. It
may or may not be a good idea to migrate the remaining Python files to
json in the future.

* Fix id lemmas.json

The contents of this file were originally just copied from the Python
source, but that used single quotes, so it had to be properly converted
to json first.

* Add .json.gz to gitignore

This covers the json.gz files built as part of distribution.

* Add language data gzip to build process

Currently this gzip data on every build; it works, but it should be
changed to only gzip when the source file has been updated.

* Remove Danish lemmatizer.py

Missed this when I added the json.

* Update to match latest explosion/srsly#9

The way gzipped json is loaded/saved in srsly changed a bit.

* Only compress language data if necessary

If a .json.gz file exists and is newer than the corresponding json file,
it's not recompressed.

* Move en/el language data to json

This only affected files >500kb, which was nouns for both languages and
the generic lookup table for English.

* Remove empty files in Norwegian tokenizer

It's unclear why, but the Norwegian (nb) tokenizer had empty files for
adj/adv/noun/verb lemmas. This may have been a result of copying the
structure of the English lemmatizer.

This removed the files, but still creates the empty sets in the
lemmatizer. That may not actually be necessary.

* Remove dubious entries in English lookup.json

" furthest" and " skilled" - both prefixed with a space - were in the
English lookup table. That seems obviously wrong so I have removed them.

* Fix small issues with en/fr lemmatizers

The en tokenizer was including the removed _nouns.py file, so that's
removed.

The fr tokenizer is unusual in that it has a lemmatizer directory with
both __init__.py and lemmatizer.py. lemmatizer.py had not been converted
to load the json language data, so that was fixed.

* Auto-format

* Auto-format

* Update srsly pin

* Consistently use pathlib paths

MANIFEST.in

@@ -3,3 +3,5 @@ include LICENSE
 include README.md
 include pyproject.toml
 include bin/spacy
+recursive-exclude spacy/lang *.json
+recursive-include spacy/lang *.json.gz

requirements.txt

@@ -5,7 +5,7 @@ thinc>=7.0.8,<7.1.0
 blis>=0.2.2,<0.3.0
 murmurhash>=0.28.0,<1.1.0
 wasabi>=0.2.0,<1.1.0
-srsly>=0.0.6,<1.1.0
+srsly>=0.1.0,<1.1.0
 # Third party dependencies
 numpy>=1.15.0
 requests>=2.13.0,<3.0.0

setup.py

@@ -132,6 +132,23 @@ def generate_cython(root, source):
         raise RuntimeError("Running cythonize failed")
 
 
+def gzip_language_data(root, source):
+    print("Compressing language data")
+    import srsly
+    from pathlib import Path
+
+    base = Path(root) / source
+    for jsonfile in base.glob("**/*.json"):
+        outfile = jsonfile.with_suffix(jsonfile.suffix + ".gz")
+        if outfile.is_file() and outfile.stat().st_ctime > jsonfile.stat().st_ctime:
+            # If the gz is newer it doesn't need updating
+            print("Skipping {}, already compressed".format(jsonfile))
+            continue
+        data = srsly.read_json(jsonfile)
+        srsly.write_gzip_json(outfile, data)
+        print("Compressed {}".format(jsonfile))
+
+
 def is_source_release(path):
     return os.path.exists(os.path.join(path, "PKG-INFO"))
 
@@ -207,6 +224,7 @@ def setup_package():
 
         if not is_source_release(root):
             generate_cython(root, "spacy")
+            gzip_language_data(root, "spacy/lang")
 
         setup(
             name="spacy",
@@ -233,7 +251,7 @@ def setup_package():
                 "plac<1.0.0,>=0.9.6",
                 "requests>=2.13.0,<3.0.0",
                 "wasabi>=0.2.0,<1.1.0",
-                "srsly>=0.0.6,<1.1.0",
+                "srsly>=0.1.0,<1.1.0",
                 'pathlib==1.0.1; python_version < "3.4"',
             ],
             setup_requires=["wheel"],

spacy/lang/ca/__init__.py

@@ -1,16 +1,17 @@
 # coding: utf8
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from .tokenizer_exceptions import TOKENIZER_EXCEPTIONS
 from .stop_words import STOP_WORDS
 from .lex_attrs import LEX_ATTRS
-from .lemmatizer import LOOKUP
 
 from ..tokenizer_exceptions import BASE_EXCEPTIONS
 from ..norm_exceptions import BASE_NORMS
 from ...language import Language
 from ...attrs import LANG, NORM
-from ...util import update_exc, add_lookups
+from ...util import update_exc, add_lookups, load_language_data
 
 from .punctuation import TOKENIZER_INFIXES
 
@@ -24,7 +25,8 @@ class CatalanDefaults(Language.Defaults):
     lex_attr_getters.update(LEX_ATTRS)
     tokenizer_exceptions = update_exc(BASE_EXCEPTIONS, TOKENIZER_EXCEPTIONS)
     stop_words = STOP_WORDS
-    lemma_lookup = LOOKUP
+    lemma_path = Path(__file__).parent / "lemmas.json"
+    lemma_lookup = load_language_data(lemma_path)
     infixes = TOKENIZER_INFIXES
 
 

spacy/lang/ca/lemmas.json

@@ -1,8 +1,4 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-
-LOOKUP = {
+{
     "as": "a",
     "àbacs": "àbac",
     "abacàs": "abacà",

spacy/lang/da/__init__.py

@@ -1,6 +1,8 @@
 # coding: utf8
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from .tokenizer_exceptions import TOKENIZER_EXCEPTIONS
 from .norm_exceptions import NORM_EXCEPTIONS
 from .punctuation import TOKENIZER_INFIXES, TOKENIZER_SUFFIXES
@@ -8,13 +10,12 @@ from .stop_words import STOP_WORDS
 from .lex_attrs import LEX_ATTRS
 from .morph_rules import MORPH_RULES
 from ..tag_map import TAG_MAP
-from .lemmatizer import LOOKUP
 
 from ..tokenizer_exceptions import BASE_EXCEPTIONS
 from ..norm_exceptions import BASE_NORMS
 from ...language import Language
 from ...attrs import LANG, NORM
-from ...util import update_exc, add_lookups
+from ...util import update_exc, add_lookups, load_language_data
 
 
 class DanishDefaults(Language.Defaults):
@@ -30,7 +31,8 @@ class DanishDefaults(Language.Defaults):
     suffixes = TOKENIZER_SUFFIXES
     tag_map = TAG_MAP
     stop_words = STOP_WORDS
-    lemma_lookup = LOOKUP
+    lemma_path = Path(__file__).parent / "lemmas.json"
+    lemma_lookup = load_language_data(lemma_path)
 
 
 class Danish(Language):

spacy/lang/da/lemmas.json

@@ -1,7 +1,4 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-LOOKUP = {
+{
     "3d-billede": "3d-billede",
     "3d-billeder": "3d-billede",
     "3d-billederne": "3d-billede",

spacy/lang/de/__init__.py

@@ -1,19 +1,20 @@
 # coding: utf8
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from .tokenizer_exceptions import TOKENIZER_EXCEPTIONS
 from .norm_exceptions import NORM_EXCEPTIONS
 from .punctuation import TOKENIZER_INFIXES
 from .tag_map import TAG_MAP
 from .stop_words import STOP_WORDS
-from .lemmatizer import LOOKUP
 from .syntax_iterators import SYNTAX_ITERATORS
 
 from ..tokenizer_exceptions import BASE_EXCEPTIONS
 from ..norm_exceptions import BASE_NORMS
 from ...language import Language
 from ...attrs import LANG, NORM
-from ...util import update_exc, add_lookups
+from ...util import update_exc, add_lookups, load_language_data
 
 
 class GermanDefaults(Language.Defaults):
@@ -27,7 +28,8 @@ class GermanDefaults(Language.Defaults):
     tag_map = TAG_MAP
     stop_words = STOP_WORDS
     syntax_iterators = SYNTAX_ITERATORS
-    lemma_lookup = LOOKUP
+    lemma_path = Path(__file__).parent / "lemmas.json"
+    lemma_lookup = load_language_data(lemma_path)
 
 
 class German(Language):

spacy/lang/de/lemmas.json

@@ -1,7 +1,4 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-LOOKUP = {
+{
     "Ab": "ab",
     "Aber": "aber",
     "ABMs": "ABM",

spacy/lang/el/lemmatizer/__init__.py

@@ -1,16 +1,20 @@
 # coding: utf8
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from ._adjectives import ADJECTIVES
 from ._adjectives_irreg import ADJECTIVES_IRREG
 from ._adverbs import ADVERBS
 from ._nouns_irreg import NOUNS_IRREG
 from ._dets_irreg import DETS_IRREG
 from ._verbs_irreg import VERBS_IRREG
-from ._nouns import NOUNS
 from ._verbs import VERBS
 from ._lemma_rules import ADJECTIVE_RULES, NOUN_RULES, VERB_RULES, PUNCT_RULES
 
+from ....util import load_language_data
+
+NOUNS = load_language_data(Path(__file__).parent / "_nouns.json")
 
 LEMMA_INDEX = {"adj": ADJECTIVES, "adv": ADVERBS, "noun": NOUNS, "verb": VERBS}
 

spacy/lang/en/lemmatizer/__init__.py

@@ -1,17 +1,21 @@
 # coding: utf8
 from __future__ import unicode_literals
 
-from .lookup import LOOKUP  # noqa: F401
+from pathlib import Path
+
 from ._adjectives import ADJECTIVES
 from ._adjectives_irreg import ADJECTIVES_IRREG
 from ._adverbs import ADVERBS
 from ._adverbs_irreg import ADVERBS_IRREG
-from ._nouns import NOUNS
 from ._nouns_irreg import NOUNS_IRREG
 from ._verbs import VERBS
 from ._verbs_irreg import VERBS_IRREG
 from ._lemma_rules import ADJECTIVE_RULES, NOUN_RULES, VERB_RULES, PUNCT_RULES
 
+from ....util import load_language_data
+
+LOOKUP = load_language_data(Path(__file__).parent / "lookup.json")
+NOUNS = load_language_data(Path(__file__).parent / "_nouns.json")
 
 LEMMA_INDEX = {"adj": ADJECTIVES, "adv": ADVERBS, "noun": NOUNS, "verb": VERBS}
 

spacy/lang/en/lemmatizer/lookup.json

@@ -1,9 +1,4 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-LOOKUP = {
-    " furtherst": "further",
-    " skilled": "skill",
+{
     "'cause": "because",
     "'d": "would",
     "'em": "them",

spacy/lang/es/__init__.py

@@ -1,18 +1,19 @@
 # coding: utf8
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from .tokenizer_exceptions import TOKENIZER_EXCEPTIONS
 from .tag_map import TAG_MAP
 from .stop_words import STOP_WORDS
 from .lex_attrs import LEX_ATTRS
-from .lemmatizer import LOOKUP
 from .syntax_iterators import SYNTAX_ITERATORS
 
 from ..tokenizer_exceptions import BASE_EXCEPTIONS
 from ..norm_exceptions import BASE_NORMS
 from ...language import Language
 from ...attrs import LANG, NORM
-from ...util import update_exc, add_lookups
+from ...util import update_exc, add_lookups, load_language_data
 
 
 class SpanishDefaults(Language.Defaults):
@@ -26,7 +27,8 @@ class SpanishDefaults(Language.Defaults):
     tag_map = TAG_MAP
     stop_words = STOP_WORDS
     syntax_iterators = SYNTAX_ITERATORS
-    lemma_lookup = LOOKUP
+    lemma_path = Path(__file__).parent / "lemmas.json"
+    lemma_lookup = load_language_data(lemma_path)
 
 
 class Spanish(Language):

spacy/lang/es/lemmas.json

@@ -1,8 +1,4 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-
-LOOKUP = {
+{
     "aba": "abar",
     "ababa": "abar",
     "ababais": "abar",

spacy/lang/fr/lemmatizer/__init__.py

@@ -1,9 +1,9 @@
 # coding: utf8
 from __future__ import unicode_literals
 
-from .lookup import LOOKUP
+from pathlib import Path
+
 from ._adjectives import ADJECTIVES
-from ._adjectives_irreg import ADJECTIVES_IRREG
 from ._adp_irreg import ADP_IRREG
 from ._adverbs import ADVERBS
 from ._auxiliary_verbs_irreg import AUXILIARY_VERBS_IRREG
@@ -15,8 +15,14 @@ from ._nouns_irreg import NOUNS_IRREG
 from ._pronouns_irreg import PRONOUNS_IRREG
 from ._sconj_irreg import SCONJ_IRREG
 from ._verbs import VERBS
-from ._verbs_irreg import VERBS_IRREG
 
+from ....util import load_language_data
+
+BASE_PATH =  Path(__file__).parent
+
+LOOKUP = load_language_data(BASE_PATH / 'lookup.json')
+VERBS_IRREG = load_language_data(BASE_PATH / '_verbs_irreg.json')
+ADJECTIVES_IRREG = load_language_data(BASE_PATH / '_adjectives_irreg.json')
 
 LEMMA_INDEX = {'adj': ADJECTIVES, 'adv': ADVERBS, 'noun': NOUNS, 'verb': VERBS}
 

spacy/lang/fr/lemmatizer/lemmatizer.py

@@ -1,9 +1,13 @@
 # coding: utf8
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from ....symbols import POS, NOUN, VERB, ADJ, ADV, PRON, DET, AUX, PUNCT, ADP, SCONJ, CCONJ
 from ....symbols import VerbForm_inf, VerbForm_none, Number_sing, Degree_pos
-from .lookup import LOOKUP
+from ....util import load_language_data
+
+LOOKUP = load_language_data(Path(__file__).parent / 'lookup.json')
 
 '''
 French language lemmatizer applies the default rule based lemmatization

spacy/lang/hu/__init__.py

@@ -1,16 +1,17 @@
 # coding: utf8
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from .tokenizer_exceptions import TOKENIZER_EXCEPTIONS, TOKEN_MATCH
 from .punctuation import TOKENIZER_PREFIXES, TOKENIZER_SUFFIXES, TOKENIZER_INFIXES
 from .stop_words import STOP_WORDS
-from .lemmatizer import LOOKUP
 
 from ..tokenizer_exceptions import BASE_EXCEPTIONS
 from ..norm_exceptions import BASE_NORMS
 from ...language import Language
 from ...attrs import LANG, NORM
-from ...util import update_exc, add_lookups
+from ...util import update_exc, add_lookups, load_language_data
 
 
 class HungarianDefaults(Language.Defaults):
@@ -25,7 +26,8 @@ class HungarianDefaults(Language.Defaults):
     suffixes = TOKENIZER_SUFFIXES
     infixes = TOKENIZER_INFIXES
     token_match = TOKEN_MATCH
-    lemma_lookup = LOOKUP
+    lemma_path = Path(__file__).parent / "lemmas.json"
+    lemma_lookup = load_language_data(lemma_path)
 
 
 class Hungarian(Language):

spacy/lang/hu/lemmas.json

@@ -1,8 +1,4 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-
-LOOKUP = {
+{
     "Aaronsonnak": "Aaronson",
     "Aaronsont": "Aaronson",
     "Abbáziában": "Abbázia",

spacy/lang/id/__init__.py

@@ -1,11 +1,12 @@
 # coding: utf8
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from .stop_words import STOP_WORDS
 from .punctuation import TOKENIZER_SUFFIXES, TOKENIZER_PREFIXES, TOKENIZER_INFIXES
 from .tokenizer_exceptions import TOKENIZER_EXCEPTIONS
 from .norm_exceptions import NORM_EXCEPTIONS
-from .lemmatizer import LOOKUP
 from .lex_attrs import LEX_ATTRS
 from .syntax_iterators import SYNTAX_ITERATORS
 from .tag_map import TAG_MAP
@@ -14,7 +15,7 @@ from ..tokenizer_exceptions import BASE_EXCEPTIONS
 from ..norm_exceptions import BASE_NORMS
 from ...language import Language
 from ...attrs import LANG, NORM
-from ...util import update_exc, add_lookups
+from ...util import update_exc, add_lookups, load_language_data
 
 
 class IndonesianDefaults(Language.Defaults):
@@ -30,7 +31,8 @@ class IndonesianDefaults(Language.Defaults):
     suffixes = TOKENIZER_SUFFIXES
     infixes = TOKENIZER_INFIXES
     syntax_iterators = SYNTAX_ITERATORS
-    lemma_lookup = LOOKUP
+    lemma_path = Path(__file__).parent / "lemmas.json"
+    lemma_lookup = load_language_data(lemma_path)
     tag_map = TAG_MAP
 
 

spacy/lang/it/__init__.py

@@ -1,8 +1,9 @@
 # coding: utf8
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from .stop_words import STOP_WORDS
-from .lemmatizer import LOOKUP
 from .tag_map import TAG_MAP
 from .tokenizer_exceptions import TOKENIZER_EXCEPTIONS
 
@@ -10,7 +11,7 @@ from ..tokenizer_exceptions import BASE_EXCEPTIONS
 from ..norm_exceptions import BASE_NORMS
 from ...language import Language
 from ...attrs import LANG, NORM
-from ...util import update_exc, add_lookups
+from ...util import update_exc, add_lookups, load_language_data
 
 from .punctuation import TOKENIZER_INFIXES
 
@@ -23,7 +24,8 @@ class ItalianDefaults(Language.Defaults):
     )
     tokenizer_exceptions = update_exc(BASE_EXCEPTIONS, TOKENIZER_EXCEPTIONS)
     stop_words = STOP_WORDS
-    lemma_lookup = LOOKUP
+    lemma_path = Path(__file__).parent / "lemmas.json"
+    lemma_lookup = load_language_data(lemma_path)
     tag_map = TAG_MAP
     infixes = TOKENIZER_INFIXES
 

spacy/lang/it/lemmas.json

@@ -1,8 +1,4 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-
-LOOKUP = {
+{
     "'ndranghete": "'ndrangheta",
     "'ndrine": "'ndrina",
     "a-storica": "a-storico",

spacy/lang/lt/__init__.py

@@ -1,18 +1,19 @@
 # coding: utf8
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from .tokenizer_exceptions import TOKENIZER_EXCEPTIONS
 from .stop_words import STOP_WORDS
 from .lex_attrs import LEX_ATTRS
 from .tag_map import TAG_MAP
-from .lemmatizer import LOOKUP
 from .morph_rules import MORPH_RULES
 
 from ..tokenizer_exceptions import BASE_EXCEPTIONS
 from ..norm_exceptions import BASE_NORMS
 from ...language import Language
 from ...attrs import LANG, NORM
-from ...util import update_exc, add_lookups
+from ...util import update_exc, add_lookups, load_language_data
 
 
 def _return_lt(_):
@@ -31,7 +32,8 @@ class LithuanianDefaults(Language.Defaults):
     stop_words = STOP_WORDS
     tag_map = TAG_MAP
     morph_rules = MORPH_RULES
-    lemma_lookup = LOOKUP
+    lemma_path = Path(__file__).parent / "lemmas.json"
+    lemma_lookup = load_language_data(lemma_path)
 
 
 class Lithuanian(Language):

spacy/lang/lt/lemmas.json

@@ -1,8 +1,4 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-
-LOOKUP = {
+{
     "Abchazas": "abchazas",
     "Abchazų": "abchazas",
     "abchazų": "abchazas",

spacy/lang/nb/lemmatizer/__init__.py

@@ -3,24 +3,27 @@
 # structure copied from the English lemmatizer
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from .lookup import LOOKUP
-from ._adjectives_wordforms import ADJECTIVES_WORDFORMS
 from ._adverbs_wordforms import ADVERBS_WORDFORMS
-from ._nouns_wordforms import NOUNS_WORDFORMS
-from ._verbs_wordforms import VERBS_WORDFORMS
 from ._lemma_rules import ADJECTIVE_RULES, NOUN_RULES, VERB_RULES, PUNCT_RULES
-from ._verbs import VERBS
-from ._nouns import NOUNS
-from ._adjectives import ADJECTIVES
-from ._adverbs import ADVERBS
+from ....util import load_language_data
+
+ADJECTIVES = set()
+ADVERBS = set()
+NOUNS = set()
+VERBS = set()
 
 LEMMA_INDEX = {"adj": ADJECTIVES, "adv": ADVERBS, "noun": NOUNS, "verb": VERBS}
 
+BASE_PATH = Path(__file__).parent 
+
 LEMMA_EXC = {
-    "adj": ADJECTIVES_WORDFORMS,
+    "adj": load_language_data(BASE_PATH / '_adjectives_wordforms.json'),
     "adv": ADVERBS_WORDFORMS,
-    "noun": NOUNS_WORDFORMS,
-    "verb": VERBS_WORDFORMS,
+    "noun": load_language_data(BASE_PATH / '_nouns_wordforms.json'),
+    "verb": load_language_data(BASE_PATH / '_verbs_wordforms.json'),
 }
 
 LEMMA_RULES = {
@@ -29,3 +32,12 @@ LEMMA_RULES = {
     "verb": VERB_RULES,
     "punct": PUNCT_RULES,
 }
+
+# Note on noun wordforms / lemmas:
+# All wordforms are extracted from Norsk Ordbank in Norwegian Bokmål 2005, updated 20180627
+# (CLARINO NB - Språkbanken), Nasjonalbiblioteket, Norway:
+# https://www.nb.no/sprakbanken/show?serial=oai%3Anb.no%3Asbr-5&lang=en
+# License:
+# Creative_Commons-BY (CC-BY) (https://creativecommons.org/licenses/by/4.0/)
+
+

spacy/lang/nb/lemmatizer/_adjectives.py

@@ -1,8 +0,0 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-
-ADJECTIVES = set(
-    """
-	""".split()
-)

spacy/lang/nb/lemmatizer/_adverbs.py

@@ -1,8 +0,0 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-
-ADVERBS = set(
-    """
-""".split()
-)

spacy/lang/nb/lemmatizer/_nouns.py

@@ -1,8 +0,0 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-
-NOUNS = set(
-    """
-""".split()
-)

spacy/lang/nb/lemmatizer/_verbs.py

@@ -1,8 +0,0 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-
-VERBS = set(
-    """
-""".split()
-)

spacy/lang/nl/lemmatizer/__init__.py

@@ -1,6 +1,8 @@
 # coding: utf8
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from ._verbs_irreg import VERBS_IRREG
 from ._nouns_irreg import NOUNS_IRREG
 from ._adjectives_irreg import ADJECTIVES_IRREG
@@ -11,16 +13,20 @@ from ._determiners_irreg import DETERMINERS_IRREG
 from ._pronouns_irreg import PRONOUNS_IRREG
 
 from ._verbs import VERBS
-from ._nouns import NOUNS
 from ._adjectives import ADJECTIVES
 
 from ._adpositions import ADPOSITIONS
 from ._determiners import DETERMINERS
 
-from .lookup import LOOKUP
 from ._lemma_rules import RULES
 from .lemmatizer import DutchLemmatizer
 
+from ....util import load_language_data
+
+BASE_PATH = Path(__file__).parent
+
+LOOKUP = load_language_data(BASE_PATH / "lookup.json")
+NOUNS = load_language_data(BASE_PATH / "_nouns.json")
 
 LEMMA_INDEX = {
     "adj": ADJECTIVES,

spacy/lang/pt/__init__.py

@@ -1,10 +1,11 @@
 # coding: utf8
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from .tokenizer_exceptions import TOKENIZER_EXCEPTIONS
 from .stop_words import STOP_WORDS
 from .lex_attrs import LEX_ATTRS
-from .lemmatizer import LOOKUP
 from .tag_map import TAG_MAP
 from .norm_exceptions import NORM_EXCEPTIONS
 
@@ -13,7 +14,7 @@ from .punctuation import TOKENIZER_INFIXES, TOKENIZER_PREFIXES
 from ..norm_exceptions import BASE_NORMS
 from ...language import Language
 from ...attrs import LANG, NORM
-from ...util import update_exc, add_lookups
+from ...util import update_exc, add_lookups, load_language_data
 
 
 class PortugueseDefaults(Language.Defaults):
@@ -25,7 +26,8 @@ class PortugueseDefaults(Language.Defaults):
     lex_attr_getters.update(LEX_ATTRS)
     tokenizer_exceptions = update_exc(BASE_EXCEPTIONS, TOKENIZER_EXCEPTIONS)
     stop_words = STOP_WORDS
-    lemma_lookup = LOOKUP
+    lemma_path = Path(__file__).parent / "lemmas.json"
+    lemma_lookup = load_language_data(lemma_path)
     tag_map = TAG_MAP
     infixes = TOKENIZER_INFIXES
     prefixes = TOKENIZER_PREFIXES

spacy/lang/pt/lemmas.json

@@ -1,8 +1,4 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-
-LOOKUP = {
+{
     "Abris": "abril",
     "Agostos": "agosto",
     "Cérberos": "cérbero",

spacy/lang/ro/__init__.py

@@ -1,15 +1,20 @@
 # coding: utf8
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from .tokenizer_exceptions import TOKENIZER_EXCEPTIONS
 from .stop_words import STOP_WORDS
-from .lemmatizer import LOOKUP
 
 from ..tokenizer_exceptions import BASE_EXCEPTIONS
 from ..norm_exceptions import BASE_NORMS
 from ...language import Language
 from ...attrs import LANG, NORM
-from ...util import update_exc, add_lookups
+from ...util import update_exc, add_lookups, load_language_data
+
+# Lemma data note:
+# Original pairs downloaded from http://www.lexiconista.com/datasets/lemmatization/
+# Replaced characters using cedillas with the correct ones (ș and ț)
 
 
 class RomanianDefaults(Language.Defaults):
@@ -20,7 +25,8 @@ class RomanianDefaults(Language.Defaults):
     )
     tokenizer_exceptions = update_exc(BASE_EXCEPTIONS, TOKENIZER_EXCEPTIONS)
     stop_words = STOP_WORDS
-    lemma_lookup = LOOKUP
+    lemma_path = Path(__file__).parent / "lemmas.json"
+    lemma_lookup = load_language_data(lemma_path)
 
 
 class Romanian(Language):

spacy/lang/ro/lemmas.json

@@ -1,10 +1,4 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-# Original pairs downloaded from http://www.lexiconista.com/datasets/lemmatization/
-# Replaced characters using cedillas with the correct ones (ș and ț)
-
-LOOKUP = {
+{
     "abale": "aba",
     "abalei": "aba",
     "abalele": "aba",

spacy/lang/sv/lemmatizer/__init__.py

@@ -1,8 +1,12 @@
 # coding: utf8
 from __future__ import unicode_literals
 
-from .lookup import LOOKUP  # noqa: F401
+from pathlib import Path
 
+from ....util import load_language_data
+
+lemma_path = Path(__file__).parent / "lookup.json"
+LOOKUP = load_language_data(lemma_path)
 
 LEMMA_RULES = {
     "noun": [

spacy/lang/sv/lemmatizer/lookup.json

@@ -1,8 +1,4 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-
-LOOKUP = {
+{
     "A-bombens": "A-bomb",
     "A-bomberna": "A-bomb",
     "A-bombers": "A-bomb",

spacy/lang/tr/__init__.py

@@ -1,15 +1,19 @@
 # coding: utf8
 from __future__ import unicode_literals
 
+from pathlib import Path
+
 from .tokenizer_exceptions import TOKENIZER_EXCEPTIONS
-from .lemmatizer import LOOKUP
 from .stop_words import STOP_WORDS
 
 from ..tokenizer_exceptions import BASE_EXCEPTIONS
 from ..norm_exceptions import BASE_NORMS
 from ...language import Language
 from ...attrs import LANG, NORM
-from ...util import update_exc, add_lookups
+from ...util import update_exc, add_lookups, load_language_data
+
+# Lemma data source:
+# http://st2.zargan.com/duyuru/Zargan_Linguistic_Resources_for_Turkish.html - Bilgin, O. (2016). Biçimbilimsel Bakımdan Karmaşık Türkçe Kelimelerin İşlenmesinde Frekans Etkileri (yayınlanmamış yüksek lisans tezi). Boğaziçi Üniversitesi, İstanbul. Erişim: http://st2.zargan.com/public/resources/turkish/frequency_effects_in_turkish.pdf
 
 
 class TurkishDefaults(Language.Defaults):
@@ -20,7 +24,8 @@ class TurkishDefaults(Language.Defaults):
     )
     tokenizer_exceptions = update_exc(BASE_EXCEPTIONS, TOKENIZER_EXCEPTIONS)
     stop_words = STOP_WORDS
-    lemma_lookup = LOOKUP
+    lemma_path = Path(__file__).parent / "lemmas.json"
+    lemma_lookup = load_language_data(lemma_path)
 
 
 class Turkish(Language):

spacy/lang/tr/lemmas.json

@@ -1,10 +1,4 @@
-# coding: utf8
-from __future__ import unicode_literals
-
-# Source: http://st2.zargan.com/duyuru/Zargan_Linguistic_Resources_for_Turkish.html - Bilgin, O. (2016). Biçimbilimsel Bakımdan Karmaşık Türkçe Kelimelerin İşlenmesinde Frekans Etkileri (yayınlanmamış yüksek lisans tezi). Boğaziçi Üniversitesi, İstanbul. Erişim: http://st2.zargan.com/public/resources/turkish/frequency_effects_in_turkish.pdf
-
-
-LOOKUP = {
+{
     "çöğürü": "çöğür",
     "çöğüründen": "çöğür",
     "çöğür": "çöğür",

spacy/util.py

@@ -117,6 +117,19 @@ def ensure_path(path):
         return path
 
 
+def load_language_data(path):
+    """Load JSON language data using the given path as a base.
+
+    If the provided path isn't present, will attempt to load a gzipped version
+    before giving up.
+    """
+
+    try:
+        return srsly.read_json(path)
+    except FileNotFoundError:
+        return srsly.read_gzip_json(path + ".gz")
+
+
 def load_model(name, **overrides):
     """Load a model from a shortcut link, package or data path.