explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2025-05-30 02:33:07 +03:00
Code Issues Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'origin/fix/types' into fix/types

Browse Source
This commit is contained in:
svlandeg 2023-12-07 13:43:59 +01:00
commit a2c9eeca0b
32 changed files with 734 additions and 58 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
.github/FUNDING.yml vendored Normal file
View File

@ -0,0 +1 @@
custom: https://explosion.ai/merch

2
README.md
View File

@ -46,6 +46,7 @@ open-source software, released under the
| 📺 **[Videos]** | Our YouTube channel with video tutorials, talks and more. |
| 🛠 **[Changelog]** | Changes and version history. |
| 💝 **[Contribute]** | How to contribute to the spaCy project and code base. |
| 👕 **[Swag]** | Support us and our work with unique, custom-designed swag! |
| <a href="https://explosion.ai/spacy-tailored-pipelines"><img src="https://user-images.githubusercontent.com/13643239/152853098-1c761611-ccb0-4ec6-9066-b234552831fe.png" width="125" alt="spaCy Tailored Pipelines"/></a> | Get a custom spaCy pipeline, tailor-made for your NLP problem by spaCy's core developers. Streamlined, production-ready, predictable and maintainable. Start by completing our 5-minute questionnaire to tell us what you need and we'll be in touch! **[Learn more &rarr;](https://explosion.ai/spacy-tailored-pipelines)** |
| <a href="https://explosion.ai/spacy-tailored-analysis"><img src="https://user-images.githubusercontent.com/1019791/206151300-b00cd189-e503-4797-aa1e-1bb6344062c5.png" width="125" alt="spaCy Tailored Pipelines"/></a> | Bespoke advice for problem solving, strategy and analysis for applied NLP projects. Services include data strategy, code reviews, pipeline design and annotation coaching. Curious? Fill in our 5-minute questionnaire to tell us what you need and we'll be in touch! **[Learn more &rarr;](https://explosion.ai/spacy-tailored-analysis)** |
@ -61,6 +62,7 @@ open-source software, released under the
[project templates]: https://github.com/explosion/projects
[changelog]: https://spacy.io/usage#changelog
[contribute]: https://github.com/explosion/spaCy/blob/master/CONTRIBUTING.md
[swag]: https://explosion.ai/merch
## 💬 Where to ask questions

17
spacy/cli/templates/quickstart_training.jinja
View File

@ -271,8 +271,9 @@ grad_factor = 1.0
@layers = "reduce_mean.v1"
[components.textcat.model.linear_model]
@architectures = "spacy.TextCatBOW.v2"
@architectures = "spacy.TextCatBOW.v3"
exclusive_classes = true
length = 262144
ngram_size = 1
no_output_layer = false
@ -308,8 +309,9 @@ grad_factor = 1.0
@layers = "reduce_mean.v1"
[components.textcat_multilabel.model.linear_model]
@architectures = "spacy.TextCatBOW.v2"
@architectures = "spacy.TextCatBOW.v3"
exclusive_classes = false
length = 262144
ngram_size = 1
no_output_layer = false
@ -542,14 +544,15 @@ nO = null
width = ${components.tok2vec.model.encode.width}
[components.textcat.model.linear_model]
@architectures = "spacy.TextCatBOW.v2"
@architectures = "spacy.TextCatBOW.v3"
exclusive_classes = true
length = 262144
ngram_size = 1
no_output_layer = false
{% else -%}
[components.textcat.model]
@architectures = "spacy.TextCatBOW.v2"
@architectures = "spacy.TextCatBOW.v3"
exclusive_classes = true
ngram_size = 1
no_output_layer = false
@ -570,15 +573,17 @@ nO = null
width = ${components.tok2vec.model.encode.width}
[components.textcat_multilabel.model.linear_model]
@architectures = "spacy.TextCatBOW.v2"
@architectures = "spacy.TextCatBOW.v3"
exclusive_classes = false
length = 262144
ngram_size = 1
no_output_layer = false
{% else -%}
[components.textcat_multilabel.model]
@architectures = "spacy.TextCatBOW.v2"
@architectures = "spacy.TextCatBOW.v3"
exclusive_classes = false
length = 262144
ngram_size = 1
no_output_layer = false
{%- endif %}

1
spacy/errors.py
View File

@ -983,6 +983,7 @@ class Errors(metaclass=ErrorsWithCodes):
"predicted docs when training {component}.")
E1055 = ("The 'replace_listener' callback expects {num_params} parameters, "
"but only callbacks with one or three parameters are supported")
E1056 = ("The `TextCatBOW` architecture expects a length of at least 1, was {length}.")
# Deprecated model shortcuts, only used in errors and warnings

18
spacy/lang/fo/__init__.py Normal file
View File

@ -0,0 +1,18 @@
from ...language import BaseDefaults, Language
from ..punctuation import TOKENIZER_INFIXES, TOKENIZER_PREFIXES, TOKENIZER_SUFFIXES
from .tokenizer_exceptions import TOKENIZER_EXCEPTIONS
class FaroeseDefaults(BaseDefaults):
tokenizer_exceptions = TOKENIZER_EXCEPTIONS
infixes = TOKENIZER_INFIXES
suffixes = TOKENIZER_SUFFIXES
prefixes = TOKENIZER_PREFIXES
class Faroese(Language):
lang = "fo"
Defaults = FaroeseDefaults
__all__ = ["Faroese"]

90
spacy/lang/fo/tokenizer_exceptions.py Normal file
View File

@ -0,0 +1,90 @@
from ...symbols import ORTH
from ...util import update_exc
from ..tokenizer_exceptions import BASE_EXCEPTIONS
_exc = {}
for orth in [
"apr.",
"aug.",
"avgr.",
"árg.",
"ávís.",
"beinl.",
"blkv.",
"blaðkv.",
"blm.",
"blaðm.",
"bls.",
"blstj.",
"blaðstj.",
"des.",
"eint.",
"febr.",
"fyrrv.",
"góðk.",
"h.m.",
"innt.",
"jan.",
"kl.",
"m.a.",
"mðr.",
"mió.",
"nr.",
"nto.",
"nov.",
"nút.",
"o.a.",
"o.a.m.",
"o.a.tíl.",
"o.fl.",
"ff.",
"o.m.a.",
"o.o.",
"o.s.fr.",
"o.tíl.",
"o.ø.",
"okt.",
"omf.",
"pst.",
"ritstj.",
"sbr.",
"sms.",
"smst.",
"smb.",
"sb.",
"sbrt.",
"sp.",
"sept.",
"spf.",
"spsk.",
"t.e.",
"t.s.",
"t.s.s.",
"tlf.",
"tel.",
"tsk.",
"t.o.v.",
"t.d.",
"uml.",
"ums.",
"uppl.",
"upprfr.",
"uppr.",
"útg.",
"útl.",
"útr.",
"vanl.",
"v.",
"v.h.",
"v.ø.o.",
"viðm.",
"viðv.",
"vm.",
"v.m.",
]:
_exc[orth] = [{ORTH: orth}]
capitalized = orth.capitalize()
_exc[capitalized] = [{ORTH: capitalized}]
TOKENIZER_EXCEPTIONS = update_exc(BASE_EXCEPTIONS, _exc)

20
spacy/lang/nn/__init__.py Normal file
View File

@ -0,0 +1,20 @@
from ...language import BaseDefaults, Language
from ..nb import SYNTAX_ITERATORS
from .punctuation import TOKENIZER_INFIXES, TOKENIZER_PREFIXES, TOKENIZER_SUFFIXES
from .tokenizer_exceptions import TOKENIZER_EXCEPTIONS
class NorwegianNynorskDefaults(BaseDefaults):
tokenizer_exceptions = TOKENIZER_EXCEPTIONS
prefixes = TOKENIZER_PREFIXES
infixes = TOKENIZER_INFIXES
suffixes = TOKENIZER_SUFFIXES
syntax_iterators = SYNTAX_ITERATORS
class NorwegianNynorsk(Language):
lang = "nn"
Defaults = NorwegianNynorskDefaults
__all__ = ["NorwegianNynorsk"]

15
spacy/lang/nn/examples.py Normal file
View File

@ -0,0 +1,15 @@
"""
Example sentences to test spaCy and its language models.
>>> from spacy.lang.nn.examples import sentences
>>> docs = nlp.pipe(sentences)
"""
# sentences taken from Omsetjingsminne frå Nynorsk pressekontor 2022 (https://www.nb.no/sprakbanken/en/resource-catalogue/oai-nb-no-sbr-80/)
sentences = [
"Konseptet går ut på at alle tre omgangar tel, alle hopparar må stille i kvalifiseringa og poengsummen skal telje.",
"Det er ein meir enn i same periode i fjor.",
"Det har lava ned enorme snømengder i store delar av Europa den siste tida.",
"Akhtar Chaudhry er ikkje innstilt på Oslo-lista til SV, men utfordrar Heikki Holmås om førsteplassen.",
]

74
spacy/lang/nn/punctuation.py Normal file
View File

@ -0,0 +1,74 @@
from ..char_classes import (
ALPHA,
ALPHA_LOWER,
ALPHA_UPPER,
CONCAT_QUOTES,
CURRENCY,
LIST_CURRENCY,
LIST_ELLIPSES,
LIST_ICONS,
LIST_PUNCT,
LIST_QUOTES,
PUNCT,
UNITS,
)
from ..punctuation import TOKENIZER_SUFFIXES
_quotes = CONCAT_QUOTES.replace("'", "")
_list_punct = [x for x in LIST_PUNCT if x != "#"]
_list_icons = [x for x in LIST_ICONS if x != "°"]
_list_icons = [x.replace("\\u00B0", "") for x in _list_icons]
_list_quotes = [x for x in LIST_QUOTES if x != "\\'"]
_prefixes = (
["§", "%", "=", "—", "–", r"\+(?![0-9])"]
+ _list_punct
+ LIST_ELLIPSES
+ LIST_QUOTES
+ LIST_CURRENCY
+ LIST_ICONS
)
_infixes = (
LIST_ELLIPSES
+ _list_icons
+ [
r"(?<=[{al}])\.(?=[{au}])".format(al=ALPHA_LOWER, au=ALPHA_UPPER),
r"(?<=[{a}])[,!?](?=[{a}])".format(a=ALPHA),
r"(?<=[{a}])[:<>=/](?=[{a}])".format(a=ALPHA),
r"(?<=[{a}]),(?=[{a}])".format(a=ALPHA),
r"(?<=[{a}])([{q}\)\]\(\[])(?=[{a}])".format(a=ALPHA, q=_quotes),
r"(?<=[{a}])--(?=[{a}])".format(a=ALPHA),
]
)
_suffixes = (
LIST_PUNCT
+ LIST_ELLIPSES
+ _list_quotes
+ _list_icons
+ ["—", "–"]
+ [
r"(?<=[0-9])\+",
r"(?<=°[FfCcKk])\.",
r"(?<=[0-9])(?:{c})".format(c=CURRENCY),
r"(?<=[0-9])(?:{u})".format(u=UNITS),
r"(?<=[{al}{e}{p}(?:{q})])\.".format(
al=ALPHA_LOWER, e=r"%²\-\+", q=_quotes, p=PUNCT
),
r"(?<=[{au}][{au}])\.".format(au=ALPHA_UPPER),
]
+ [r"(?<=[^sSxXzZ])'"]
)
_suffixes += [
suffix
for suffix in TOKENIZER_SUFFIXES
if suffix not in ["'s", "'S", "’s", "’S", r"\'"]
]
TOKENIZER_PREFIXES = _prefixes
TOKENIZER_INFIXES = _infixes
TOKENIZER_SUFFIXES = _suffixes

228
spacy/lang/nn/tokenizer_exceptions.py Normal file
View File

@ -0,0 +1,228 @@
from ...symbols import NORM, ORTH
from ...util import update_exc
from ..tokenizer_exceptions import BASE_EXCEPTIONS
_exc = {}
for exc_data in [
{ORTH: "jan.", NORM: "januar"},
{ORTH: "feb.", NORM: "februar"},
{ORTH: "mar.", NORM: "mars"},
{ORTH: "apr.", NORM: "april"},
{ORTH: "jun.", NORM: "juni"},
# note: "jul." is in the simple list below without a NORM exception
{ORTH: "aug.", NORM: "august"},
{ORTH: "sep.", NORM: "september"},
{ORTH: "okt.", NORM: "oktober"},
{ORTH: "nov.", NORM: "november"},
{ORTH: "des.", NORM: "desember"},
]:
_exc[exc_data[ORTH]] = [exc_data]
for orth in [
"Ap.",
"Aq.",
"Ca.",
"Chr.",
"Co.",
"Dr.",
"F.eks.",
"Fr.p.",
"Frp.",
"Grl.",
"Kr.",
"Kr.F.",
"Kr.F.s",
"Mr.",
"Mrs.",
"Pb.",
"Pr.",
"Sp.",
"St.",
"a.m.",
"ad.",
"adm.dir.",
"adr.",
"b.c.",
"bl.a.",
"bla.",
"bm.",
"bnr.",
"bto.",
"c.c.",
"ca.",
"cand.mag.",
"co.",
"d.d.",
"d.m.",
"d.y.",
"dept.",
"dr.",
"dr.med.",
"dr.philos.",
"dr.psychol.",
"dss.",
"dvs.",
"e.Kr.",
"e.l.",
"eg.",
"eig.",
"ekskl.",
"el.",
"et.",
"etc.",
"etg.",
"ev.",
"evt.",
"f.",
"f.Kr.",
"f.eks.",
"f.o.m.",
"fhv.",
"fk.",
"foreg.",
"fork.",
"fv.",
"fvt.",
"g.",
"gl.",
"gno.",
"gnr.",
"grl.",
"gt.",
"h.r.adv.",
"hhv.",
"hoh.",
"hr.",
"ifb.",
"ifm.",
"iht.",
"inkl.",
"istf.",
"jf.",
"jr.",
"jul.",
"juris.",
"kfr.",
"kgl.",
"kgl.res.",
"kl.",
"komm.",
"kr.",
"kst.",
"lat.",
"lø.",
"m.a.",
"m.a.o.",
"m.fl.",
"m.m.",
"m.v.",
"ma.",
"mag.art.",
"md.",
"mfl.",
"mht.",
"mill.",
"min.",
"mnd.",
"moh.",
"mrd.",
"muh.",
"mv.",
"mva.",
"n.å.",
"ndf.",
"nr.",
"nto.",
"nyno.",
"o.a.",
"o.l.",
"obl.",
"off.",
"ofl.",
"on.",
"op.",
"org.",
"osv.",
"ovf.",
"p.",
"p.a.",
"p.g.a.",
"p.m.",
"p.t.",
"pga.",
"ph.d.",
"pkt.",
"pr.",
"pst.",
"pt.",
"red.anm.",
"ref.",
"res.",
"res.kap.",
"resp.",
"rv.",
"s.",
"s.d.",
"s.k.",
"s.u.",
"s.å.",
"sen.",
"sep.",
"siviling.",
"sms.",
"snr.",
"spm.",
"sr.",
"sst.",
"st.",
"st.meld.",
"st.prp.",
"stip.",
"stk.",
"stud.",
"sv.",
"såk.",
"sø.",
"t.d.",
"t.h.",
"t.o.m.",
"t.v.",
"temp.",
"ti.",
"tils.",
"tilsv.",
"tl;dr",
"tlf.",
"to.",
"ult.",
"utg.",
"v.",
"vedk.",
"vedr.",
"vg.",
"vgs.",
"vha.",
"vit.ass.",
"vn.",
"vol.",
"vs.",
"vsa.",
"§§",
"©NTB",
"årg.",
"årh.",
]:
_exc[orth] = [{ORTH: orth}]
# Dates
for h in range(1, 31 + 1):
for period in ["."]:
_exc[f"{h}{period}"] = [{ORTH: f"{h}."}]
_custom_base_exc = {"i.": [{ORTH: "i", NORM: "i"}, {ORTH: "."}]}
_exc.update(_custom_base_exc)
TOKENIZER_EXCEPTIONS = update_exc(BASE_EXCEPTIONS, _exc)

46
spacy/ml/models/textcat.py
View File

@ -1,5 +1,5 @@
from functools import partial
from typing import List, Optional, cast
from typing import List, Optional, Tuple, cast
from thinc.api import (
Dropout,
@ -12,6 +12,7 @@ from thinc.api import (
Relu,
Softmax,
SparseLinear,
SparseLinear_v2,
chain,
clone,
concatenate,
@ -25,9 +26,10 @@ from thinc.api import (
)
from thinc.layers.chain import init as init_chain
from thinc.layers.resizable import resize_linear_weighted, resize_model
from thinc.types import Floats2d
from thinc.types import ArrayXd, Floats2d
from ...attrs import ORTH
from ...errors import Errors
from ...tokens import Doc
from ...util import registry
from ..extract_ngrams import extract_ngrams
@ -95,10 +97,48 @@ def build_bow_text_classifier(
ngram_size: int,
no_output_layer: bool,
nO: Optional[int] = None,
) -> Model[List[Doc], Floats2d]:
return _build_bow_text_classifier(
exclusive_classes=exclusive_classes,
ngram_size=ngram_size,
no_output_layer=no_output_layer,
nO=nO,
sparse_linear=SparseLinear(nO=nO),
)
@registry.architectures("spacy.TextCatBOW.v3")
def build_bow_text_classifier_v3(
exclusive_classes: bool,
ngram_size: int,
no_output_layer: bool,
length: int = 262144,
nO: Optional[int] = None,
) -> Model[List[Doc], Floats2d]:
if length < 1:
raise ValueError(Errors.E1056.format(length=length))
# Find k such that 2**(k-1) < length <= 2**k.
length = 2 ** (length - 1).bit_length()
return _build_bow_text_classifier(
exclusive_classes=exclusive_classes,
ngram_size=ngram_size,
no_output_layer=no_output_layer,
nO=nO,
sparse_linear=SparseLinear_v2(nO=nO, length=length),
)
def _build_bow_text_classifier(
exclusive_classes: bool,
ngram_size: int,
no_output_layer: bool,
sparse_linear: Model[Tuple[ArrayXd, ArrayXd, ArrayXd], ArrayXd],
nO: Optional[int] = None,
) -> Model[List[Doc], Floats2d]:
fill_defaults = {"b": 0, "W": 0}
with Model.define_operators({">>": chain}):
sparse_linear = SparseLinear(nO=nO)
output_layer = None
if not no_output_layer:
fill_defaults["b"] = NEG_VALUE

6
spacy/pipeline/textcat.py
View File

@ -36,8 +36,9 @@ maxout_pieces = 3
depth = 2
[model.linear_model]
@architectures = "spacy.TextCatBOW.v2"
@architectures = "spacy.TextCatBOW.v3"
exclusive_classes = true
length = 262144
ngram_size = 1
no_output_layer = false
"""
@ -45,8 +46,9 @@ DEFAULT_SINGLE_TEXTCAT_MODEL = Config().from_str(single_label_default_config)["m
single_label_bow_config = """
[model]
@architectures = "spacy.TextCatBOW.v2"
@architectures = "spacy.TextCatBOW.v3"
exclusive_classes = true
length = 262144
ngram_size = 1
no_output_layer = false
"""

5
spacy/pipeline/textcat_multilabel.py
View File

@ -35,8 +35,9 @@ maxout_pieces = 3
depth = 2
[model.linear_model]
@architectures = "spacy.TextCatBOW.v2"
@architectures = "spacy.TextCatBOW.v3"
exclusive_classes = false
length = 262144
ngram_size = 1
no_output_layer = false
"""
@ -44,7 +45,7 @@ DEFAULT_MULTI_TEXTCAT_MODEL = Config().from_str(multi_label_default_config)["mod
multi_label_bow_config = """
[model]
@architectures = "spacy.TextCatBOW.v2"
@architectures = "spacy.TextCatBOW.v3"
exclusive_classes = false
ngram_size = 1
no_output_layer = false

10
spacy/tests/conftest.py
View File

@ -162,6 +162,11 @@ def fi_tokenizer():
return get_lang_class("fi")().tokenizer
@pytest.fixture(scope="session")
def fo_tokenizer():
return get_lang_class("fo")().tokenizer
@pytest.fixture(scope="session")
def fr_tokenizer():
return get_lang_class("fr")().tokenizer
@ -317,6 +322,11 @@ def nl_tokenizer():
return get_lang_class("nl")().tokenizer
@pytest.fixture(scope="session")
def nn_tokenizer():
return get_lang_class("nn")().tokenizer
@pytest.fixture(scope="session")
def pl_tokenizer():
return get_lang_class("pl")().tokenizer

0
spacy/tests/lang/fo/__init__.py Normal file
View File

26
spacy/tests/lang/fo/test_tokenizer.py Normal file
View File

@ -0,0 +1,26 @@
import pytest
# examples taken from Basic LAnguage Resource Kit 1.0 for Faroese (https://maltokni.fo/en/resources) licensed with CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
# fmt: off
FO_TOKEN_EXCEPTION_TESTS = [
(
"Eftir løgtingslóg um samsýning og eftirløn landsstýrismanna v.m., skulu løgmaður og landsstýrismenn vanliga siga frá sær størv í almennari tænastu ella privatum virkjum, samtøkum ella stovnum. ",
[
"Eftir", "løgtingslóg", "um", "samsýning", "og", "eftirløn", "landsstýrismanna", "v.m.", ",", "skulu", "løgmaður", "og", "landsstýrismenn", "vanliga", "siga", "frá", "sær", "størv", "í", "almennari", "tænastu", "ella", "privatum", "virkjum", ",", "samtøkum", "ella", "stovnum", ".",
],
),
(
"Sambandsflokkurin gongur aftur við 2,7 prosentum í mun til valið í 1994, tá flokkurin fekk undirtøku frá 23,4 prosent av veljarunum.",
[
"Sambandsflokkurin", "gongur", "aftur", "við", "2,7", "prosentum", "í", "mun", "til", "valið", "í", "1994", ",", "tá", "flokkurin", "fekk", "undirtøku", "frá", "23,4", "prosent", "av", "veljarunum", ".",
],
),
]
# fmt: on
@pytest.mark.parametrize("text,expected_tokens", FO_TOKEN_EXCEPTION_TESTS)
def test_fo_tokenizer_handles_exception_cases(fo_tokenizer, text, expected_tokens):
tokens = fo_tokenizer(text)
token_list = [token.text for token in tokens if not token.is_space]
assert expected_tokens == token_list

0
spacy/tests/lang/nn/__init__.py Normal file
View File

38
spacy/tests/lang/nn/test_tokenizer.py Normal file
View File

@ -0,0 +1,38 @@
import pytest
# examples taken from Omsetjingsminne frå Nynorsk pressekontor 2022 (https://www.nb.no/sprakbanken/en/resource-catalogue/oai-nb-no-sbr-80/)
# fmt: off
NN_TOKEN_EXCEPTION_TESTS = [
(
"Målet til direktoratet er at alle skal bli tilbydd jobb i politiet så raskt som mogleg i 2014.",
[
"Målet", "til", "direktoratet", "er", "at", "alle", "skal", "bli", "tilbydd", "jobb", "i", "politiet", "så", "raskt", "som", "mogleg", "i", "2014", ".",
],
),
(
"Han ønskjer ikkje at staten skal vere med på å finansiere slik undervisning, men dette er rektor på skulen ueinig i.",
[
"Han", "ønskjer", "ikkje", "at", "staten", "skal", "vere", "med", "på", "å", "finansiere", "slik", "undervisning", ",", "men", "dette", "er", "rektor", "på", "skulen", "ueinig", "i", ".",
],
),
(
"Ifølgje China Daily vart det 8.848 meter høge fjellet flytta 3 centimeter sørvestover under jordskjelvet, som vart målt til 7,8.",
[
"Ifølgje", "China", "Daily", "vart", "det", "8.848", "meter", "høge", "fjellet", "flytta", "3", "centimeter", "sørvestover", "under", "jordskjelvet", ",", "som", "vart", "målt", "til", "7,8", ".",
],
),
(
"Brukssesongen er frå nov. til mai, med ein topp i mars.",
[
"Brukssesongen", "er", "frå", "nov.", "til", "mai", ",", "med", "ein", "topp", "i", "mars", ".",
],
),
]
# fmt: on
@pytest.mark.parametrize("text,expected_tokens", NN_TOKEN_EXCEPTION_TESTS)
def test_nn_tokenizer_handles_exception_cases(nn_tokenizer, text, expected_tokens):
tokens = nn_tokenizer(text)
token_list = [token.text for token in tokens if not token.is_space]
assert expected_tokens == token_list

2
spacy/tests/pipeline/test_pipe_factories.py
View File

@ -203,7 +203,7 @@ def test_pipe_class_component_model():
"@architectures": "spacy.TextCatEnsemble.v2",
"tok2vec": DEFAULT_TOK2VEC_MODEL,
"linear_model": {
"@architectures": "spacy.TextCatBOW.v2",
"@architectures": "spacy.TextCatBOW.v3",
"exclusive_classes": False,
"ngram_size": 1,
"no_output_layer": False,

31
spacy/tests/pipeline/test_textcat.py
View File

@ -414,7 +414,7 @@ def test_implicit_label(name, get_examples):
@pytest.mark.parametrize(
"name,textcat_config",
[
# BOW
# BOW V1
("textcat", {"@architectures": "spacy.TextCatBOW.v1", "exclusive_classes": True, "no_output_layer": False, "ngram_size": 3}),
("textcat", {"@architectures": "spacy.TextCatBOW.v1", "exclusive_classes": True, "no_output_layer": True, "ngram_size": 3}),
("textcat_multilabel", {"@architectures": "spacy.TextCatBOW.v1", "exclusive_classes": False, "no_output_layer": False, "ngram_size": 3}),
@ -451,11 +451,11 @@ def test_no_resize(name, textcat_config):
@pytest.mark.parametrize(
"name,textcat_config",
[
# BOW
("textcat", {"@architectures": "spacy.TextCatBOW.v2", "exclusive_classes": True, "no_output_layer": False, "ngram_size": 3}),
("textcat", {"@architectures": "spacy.TextCatBOW.v2", "exclusive_classes": True, "no_output_layer": True, "ngram_size": 3}),
("textcat_multilabel", {"@architectures": "spacy.TextCatBOW.v2", "exclusive_classes": False, "no_output_layer": False, "ngram_size": 3}),
("textcat_multilabel", {"@architectures": "spacy.TextCatBOW.v2", "exclusive_classes": False, "no_output_layer": True, "ngram_size": 3}),
# BOW V3
("textcat", {"@architectures": "spacy.TextCatBOW.v3", "exclusive_classes": True, "no_output_layer": False, "ngram_size": 3}),
("textcat", {"@architectures": "spacy.TextCatBOW.v3", "exclusive_classes": True, "no_output_layer": True, "ngram_size": 3}),
("textcat_multilabel", {"@architectures": "spacy.TextCatBOW.v3", "exclusive_classes": False, "no_output_layer": False, "ngram_size": 3}),
("textcat_multilabel", {"@architectures": "spacy.TextCatBOW.v3", "exclusive_classes": False, "no_output_layer": True, "ngram_size": 3}),
# CNN
("textcat", {"@architectures": "spacy.TextCatCNN.v2", "tok2vec": DEFAULT_TOK2VEC_MODEL, "exclusive_classes": True}),
("textcat_multilabel", {"@architectures": "spacy.TextCatCNN.v2", "tok2vec": DEFAULT_TOK2VEC_MODEL, "exclusive_classes": False}),
@ -480,11 +480,11 @@ def test_resize(name, textcat_config):
@pytest.mark.parametrize(
"name,textcat_config",
[
# BOW
("textcat", {"@architectures": "spacy.TextCatBOW.v2", "exclusive_classes": True, "no_output_layer": False, "ngram_size": 3}),
("textcat", {"@architectures": "spacy.TextCatBOW.v2", "exclusive_classes": True, "no_output_layer": True, "ngram_size": 3}),
("textcat_multilabel", {"@architectures": "spacy.TextCatBOW.v2", "exclusive_classes": False, "no_output_layer": False, "ngram_size": 3}),
("textcat_multilabel", {"@architectures": "spacy.TextCatBOW.v2", "exclusive_classes": False, "no_output_layer": True, "ngram_size": 3}),
# BOW v3
("textcat", {"@architectures": "spacy.TextCatBOW.v3", "exclusive_classes": True, "no_output_layer": False, "ngram_size": 3}),
("textcat", {"@architectures": "spacy.TextCatBOW.v3", "exclusive_classes": True, "no_output_layer": True, "ngram_size": 3}),
("textcat_multilabel", {"@architectures": "spacy.TextCatBOW.v3", "exclusive_classes": False, "no_output_layer": False, "ngram_size": 3}),
("textcat_multilabel", {"@architectures": "spacy.TextCatBOW.v3", "exclusive_classes": False, "no_output_layer": True, "ngram_size": 3}),
# CNN
("textcat", {"@architectures": "spacy.TextCatCNN.v2", "tok2vec": DEFAULT_TOK2VEC_MODEL, "exclusive_classes": True}),
("textcat_multilabel", {"@architectures": "spacy.TextCatCNN.v2", "tok2vec": DEFAULT_TOK2VEC_MODEL, "exclusive_classes": False}),
@ -693,9 +693,14 @@ def test_overfitting_IO_multi():
("textcat", TRAIN_DATA_SINGLE_LABEL, {"@architectures": "spacy.TextCatBOW.v2", "exclusive_classes": True, "ngram_size": 4, "no_output_layer": False}),
("textcat_multilabel", TRAIN_DATA_MULTI_LABEL, {"@architectures": "spacy.TextCatBOW.v2", "exclusive_classes": False, "ngram_size": 3, "no_output_layer": True}),
("textcat", TRAIN_DATA_SINGLE_LABEL, {"@architectures": "spacy.TextCatBOW.v2", "exclusive_classes": True, "ngram_size": 2, "no_output_layer": True}),
# BOW V3
("textcat_multilabel", TRAIN_DATA_MULTI_LABEL, {"@architectures": "spacy.TextCatBOW.v3", "exclusive_classes": False, "ngram_size": 1, "no_output_layer": False}),
("textcat", TRAIN_DATA_SINGLE_LABEL, {"@architectures": "spacy.TextCatBOW.v3", "exclusive_classes": True, "ngram_size": 4, "no_output_layer": False}),
("textcat_multilabel", TRAIN_DATA_MULTI_LABEL, {"@architectures": "spacy.TextCatBOW.v3", "exclusive_classes": False, "ngram_size": 3, "no_output_layer": True}),
("textcat", TRAIN_DATA_SINGLE_LABEL, {"@architectures": "spacy.TextCatBOW.v3", "exclusive_classes": True, "ngram_size": 2, "no_output_layer": True}),
# ENSEMBLE V2
("textcat_multilabel", TRAIN_DATA_MULTI_LABEL, {"@architectures": "spacy.TextCatEnsemble.v2", "tok2vec": DEFAULT_TOK2VEC_MODEL, "linear_model": {"@architectures": "spacy.TextCatBOW.v2", "exclusive_classes": False, "ngram_size": 1, "no_output_layer": False}}),
("textcat", TRAIN_DATA_SINGLE_LABEL, {"@architectures": "spacy.TextCatEnsemble.v2", "tok2vec": DEFAULT_TOK2VEC_MODEL, "linear_model": {"@architectures": "spacy.TextCatBOW.v2", "exclusive_classes": True, "ngram_size": 5, "no_output_layer": False}}),
("textcat_multilabel", TRAIN_DATA_MULTI_LABEL, {"@architectures": "spacy.TextCatEnsemble.v2", "tok2vec": DEFAULT_TOK2VEC_MODEL, "linear_model": {"@architectures": "spacy.TextCatBOW.v3", "exclusive_classes": False, "ngram_size": 1, "no_output_layer": False}}),
("textcat", TRAIN_DATA_SINGLE_LABEL, {"@architectures": "spacy.TextCatEnsemble.v2", "tok2vec": DEFAULT_TOK2VEC_MODEL, "linear_model": {"@architectures": "spacy.TextCatBOW.v3", "exclusive_classes": True, "ngram_size": 5, "no_output_layer": False}}),
# CNN V2
("textcat", TRAIN_DATA_SINGLE_LABEL, {"@architectures": "spacy.TextCatCNN.v2", "tok2vec": DEFAULT_TOK2VEC_MODEL, "exclusive_classes": True}),
("textcat_multilabel", TRAIN_DATA_MULTI_LABEL, {"@architectures": "spacy.TextCatCNN.v2", "tok2vec": DEFAULT_TOK2VEC_MODEL, "exclusive_classes": False}),

4
spacy/tests/test_cli_app.py
View File

@ -238,7 +238,7 @@ def test_project_push_pull(project_dir):
def test_find_function_valid():
# example of architecture in main code base
function = "spacy.TextCatBOW.v2"
function = "spacy.TextCatBOW.v3"
result = CliRunner().invoke(app, ["find-function", function, "-r", "architectures"])
assert f"Found registered function '{function}'" in result.stdout
assert "textcat.py" in result.stdout
@ -257,7 +257,7 @@ def test_find_function_valid():
def test_find_function_invalid():
# invalid registry
function = "spacy.TextCatBOW.v2"
function = "spacy.TextCatBOW.v3"
registry = "foobar"
result = CliRunner().invoke(
app, ["find-function", function, "--registry", registry]

3
spacy/tests/test_misc.py
View File

@ -376,8 +376,9 @@ def test_util_dot_section():
factory = "textcat"
[components.textcat.model]
@architectures = "spacy.TextCatBOW.v2"
@architectures = "spacy.TextCatBOW.v3"
exclusive_classes = true
length = 262144
ngram_size = 1
no_output_layer = false
"""

39
website/docs/api/architectures.mdx
View File

@ -78,16 +78,16 @@ subword features, and a
[MaxoutWindowEncoder](/api/architectures#MaxoutWindowEncoder) encoding layer
consisting of a CNN and a layer-normalized maxout activation function.
| Name | Description |
| -------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `width` | The width of the input and output. These are required to be the same, so that residual connections can be used. Recommended values are `96`, `128` or `300`. ~~int~~ |
| `depth` | The number of convolutional layers to use. Recommended values are between `2` and `8`. ~~int~~ |
| `embed_size` | The number of rows in the hash embedding tables. This can be surprisingly small, due to the use of the hash embeddings. Recommended values are between `2000` and `10000`. ~~int~~ |
| Name | Description |
| -------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `width` | The width of the input and output. These are required to be the same, so that residual connections can be used. Recommended values are `96`, `128` or `300`. ~~int~~ |
| `depth` | The number of convolutional layers to use. Recommended values are between `2` and `8`. ~~int~~ |
| `embed_size` | The number of rows in the hash embedding tables. This can be surprisingly small, due to the use of the hash embeddings. Recommended values are between `2000` and `10000`. ~~int~~ |
| `window_size` | The number of tokens on either side to concatenate during the convolutions. The receptive field of the CNN will be `depth * window_size * 2 + 1`, so a 4-layer network with a window size of `2` will be sensitive to 17 words at a time. Recommended value is `1`. ~~int~~ |
| `maxout_pieces` | The number of pieces to use in the maxout non-linearity. If `1`, the [`Mish`](https://thinc.ai/docs/api-layers#mish) non-linearity is used instead. Recommended values are `1`-`3`. ~~int~~ |
| `subword_features` | Whether to also embed subword features, specifically the prefix, suffix and word shape. This is recommended for alphabetic languages like English, but not if single-character tokens are used for a language such as Chinese. ~~bool~~ |
| `pretrained_vectors` | Whether to also use static vectors. ~~bool~~ |
| **CREATES** | The model using the architecture. ~~Model[List[Doc], List[Floats2d]]~~ |
| `maxout_pieces` | The number of pieces to use in the maxout non-linearity. If `1`, the [`Mish`](https://thinc.ai/docs/api-layers#mish) non-linearity is used instead. Recommended values are `1`-`3`. ~~int~~ |
| `subword_features` | Whether to also embed subword features, specifically the prefix, suffix and word shape. This is recommended for alphabetic languages like English, but not if single-character tokens are used for a language such as Chinese. ~~bool~~ |
| `pretrained_vectors` | Whether to also use static vectors. ~~bool~~ |
| **CREATES** | The model using the architecture. ~~Model[List[Doc], List[Floats2d]]~~ |
### spacy.Tok2VecListener.v1 {id="Tok2VecListener"}
@ -962,8 +962,9 @@ single-label use-cases where `exclusive_classes = true`, while the
> nO = null
>
> [model.linear_model]
> @architectures = "spacy.TextCatBOW.v2"
> @architectures = "spacy.TextCatBOW.v3"
> exclusive_classes = true
> length = 262144
> ngram_size = 1
> no_output_layer = false
>
@ -1057,14 +1058,15 @@ after training.
</Accordion>
### spacy.TextCatBOW.v2 {id="TextCatBOW"}
### spacy.TextCatBOW.v3 {id="TextCatBOW"}
> #### Example Config
>
> ```ini
> [model]
> @architectures = "spacy.TextCatBOW.v2"
> @architectures = "spacy.TextCatBOW.v3"
> exclusive_classes = false
> length = 262144
> ngram_size = 1
> no_output_layer = false
> nO = null
@ -1078,14 +1080,19 @@ the others, but may not be as accurate, especially if texts are short.
| `exclusive_classes` | Whether or not categories are mutually exclusive. ~~bool~~ |
| `ngram_size` | Determines the maximum length of the n-grams in the BOW model. For instance, `ngram_size=3` would give unigram, trigram and bigram features. ~~int~~ |
| `no_output_layer` | Whether or not to add an output layer to the model (`Softmax` activation if `exclusive_classes` is `True`, else `Logistic`). ~~bool~~ |
| `length` | The size of the weights vector. The length will be rounded up to the next power of two if it is not a power of two. Defaults to `262144`. ~~int~~ |
| `nO` | Output dimension, determined by the number of different labels. If not set, the [`TextCategorizer`](/api/textcategorizer) component will set it when `initialize` is called. ~~Optional[int]~~ |
| **CREATES** | The model using the architecture. ~~Model[List[Doc], Floats2d]~~ |
<Accordion title="spacy.TextCatBOW.v1 definition" spaced>
<Accordion title="Previous versions of spacy.TextCatBOW" spaced>
[TextCatBOW.v1](/api/legacy#TextCatBOW_v1) had the exact same signature, but was
not yet resizable. Since v2, new labels can be added to this component, even
after training.
- [TextCatBOW.v1](/api/legacy#TextCatBOW_v1) was not yet resizable. Since v2,
new labels can be added to this component, even after training.
- [TextCatBOW.v1](/api/legacy#TextCatBOW_v1) and
[TextCatBOW.v2](/api/legacy#TextCatBOW_v2) used an erroneous sparse linear
layer that only used a small number of the allocated parameters.
- [TextCatBOW.v1](/api/legacy#TextCatBOW_v1) and
[TextCatBOW.v2](/api/legacy#TextCatBOW_v2) did not have the `length` argument.
</Accordion>

8
website/docs/api/curatedtransformer.mdx
View File

@ -400,6 +400,14 @@ identifiers are grouped by token. Instances of this class are typically assigned
to the [`Doc._.trf_data`](/api/curatedtransformer#assigned-attributes) extension
attribute.
> #### Example
>
> ```python
> # Get the last hidden layer output for "is" (token index 1)
> doc = nlp("This is a text.")
> tensors = doc._.trf_data.last_hidden_layer_state[1]
> ```
| Name | Description |
| ----------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `all_outputs` | List of `Ragged` tensors that correspends to outputs of the different transformer layers. Each tensor element corresponds to a piece identifier's representation. ~~List[Ragged]~~ |

31
website/docs/api/legacy.mdx
View File

@ -198,7 +198,9 @@ architecture is usually less accurate than the ensemble, but runs faster.
Since `spacy.TextCatBOW.v2`, this architecture has become resizable, which means
that you can add labels to a previously trained textcat. `TextCatBOW` v1 did not
yet support that.
yet support that. Versions of this model before `spacy.TextCatBOW.v3` used an
erroneous sparse linear layer that only used a small number of the allocated
parameters.
> #### Example Config
>
@ -222,6 +224,33 @@ the others, but may not be as accurate, especially if texts are short.
| `nO` | Output dimension, determined by the number of different labels. If not set, the [`TextCategorizer`](/api/textcategorizer) component will set it when `initialize` is called. ~~Optional[int]~~ |
| **CREATES** | The model using the architecture. ~~Model[List[Doc], Floats2d]~~ |
### spacy.TextCatBOW.v2 {id="TextCatBOW"}
Versions of this model before `spacy.TextCatBOW.v3` used an erroneous sparse
linear layer that only used a small number of the allocated parameters.
> #### Example Config
>
> ```ini
> [model]
> @architectures = "spacy.TextCatBOW.v2"
> exclusive_classes = false
> ngram_size = 1
> no_output_layer = false
> nO = null
> ```
An n-gram "bag-of-words" model. This architecture should run much faster than
the others, but may not be as accurate, especially if texts are short.
| Name | Description |
| ------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| `exclusive_classes` | Whether or not categories are mutually exclusive. ~~bool~~ |
| `ngram_size` | Determines the maximum length of the n-grams in the BOW model. For instance, `ngram_size=3` would give unigram, trigram and bigram features. ~~int~~ |
| `no_output_layer` | Whether or not to add an output layer to the model (`Softmax` activation if `exclusive_classes` is `True`, else `Logistic`). ~~bool~~ |
| `nO` | Output dimension, determined by the number of different labels. If not set, the [`TextCategorizer`](/api/textcategorizer) component will set it when `initialize` is called. ~~Optional[int]~~ |
| **CREATES** | The model using the architecture. ~~Model[List[Doc], Floats2d]~~ |
### spacy.TransitionBasedParser.v1 {id="TransitionBasedParser_v1"}
Identical to

11
website/docs/api/transformer.mdx
View File

@ -397,6 +397,17 @@ are wrapped into the
by this class. Instances of this class are typically assigned to the
[`Doc._.trf_data`](/api/transformer#assigned-attributes) extension attribute.
> #### Example
>
> ```python
> # Get the last hidden layer output for "is" (token index 1)
> doc = nlp("This is a text.")
> indices = doc._.trf_data.align[1].data.flatten()
> last_hidden_state = doc._.trf_data.model_output.last_hidden_state
> dim = last_hidden_state.shape[-1]
> tensors = last_hidden_state.reshape(-1, dim)[indices]
> ```
| Name | Description |
| -------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| `tokens` | A slice of the tokens data produced by the tokenizer. This may have several fields, including the token IDs, the texts and the attention mask. See the [`transformers.BatchEncoding`](https://huggingface.co/transformers/main_classes/tokenizer.html#transformers.BatchEncoding) object for details. ~~dict~~ |

26
website/docs/models/index.mdx
View File

@ -108,12 +108,12 @@ In the `sm`/`md`/`lg` models:
#### CNN/CPU pipelines with floret vectors
The Finnish, Korean and Swedish `md` and `lg` pipelines use
[floret vectors](/usage/v3-2#vectors) instead of default vectors. If you're
running a trained pipeline on texts and working with [`Doc`](/api/doc) objects,
you shouldn't notice any difference with floret vectors. With floret vectors no
tokens are out-of-vocabulary, so [`Token.is_oov`](/api/token#attributes) will
return `False` for all tokens.
The Croatian, Finnish, Korean, Slovenian, Swedish and Ukrainian `md` and `lg`
pipelines use [floret vectors](/usage/v3-2#vectors) instead of default vectors.
If you're running a trained pipeline on texts and working with [`Doc`](/api/doc)
objects, you shouldn't notice any difference with floret vectors. With floret
vectors no tokens are out-of-vocabulary, so
[`Token.is_oov`](/api/token#attributes) will return `False` for all tokens.
If you access vectors directly for similarity comparisons, there are a few
differences because floret vectors don't include a fixed word list like the
@ -132,10 +132,20 @@ vector keys for default vectors.
### Transformer pipeline design {id="design-trf"}
In the transformer (`trf`) models, the `tagger`, `parser` and `ner` (if present)
all listen to the `transformer` component. The `attribute_ruler` and
In the transformer (`trf`) pipelines, the `tagger`, `parser` and `ner` (if
present) all listen to the `transformer` component. The `attribute_ruler` and
`lemmatizer` have the same configuration as in the CNN models.
For spaCy v3.0-v3.6, `trf` pipelines use
[`spacy-transformers`](https://github.com/explosion/spacy-transformers) and the
transformer output in `doc._.trf_data` is a
[`TransformerData`](/api/transformer#transformerdata) object.
For spaCy v3.7+, `trf` pipelines use
[`spacy-curated-transformers`](https://github.com/explosion/spacy-curated-transformers)
and `doc._.trf_data` is a
[`DocTransformerOutput`](/api/curatedtransformer#doctransformeroutput) object.
### Modifying the default pipeline {id="design-modify"}
For faster processing, you may only want to run a subset of the components in a

6
website/docs/usage/layers-architectures.mdx
View File

@ -153,8 +153,9 @@ maxout_pieces = 3
depth = 2
[components.textcat.model.linear_model]
@architectures = "spacy.TextCatBOW.v2"
@architectures = "spacy.TextCatBOW.v3"
exclusive_classes = true
length = 262144
ngram_size = 1
no_output_layer = false
```
@ -170,8 +171,9 @@ factory = "textcat"
labels = []
[components.textcat.model]
@architectures = "spacy.TextCatBOW.v2"
@architectures = "spacy.TextCatBOW.v3"
exclusive_classes = true
length = 262144
ngram_size = 1
no_output_layer = false
nO = null

3
website/docs/usage/processing-pipelines.mdx
View File

@ -1328,8 +1328,9 @@ labels = []
# This function is created and then passed to the "textcat" component as
# the argument "model"
[components.textcat.model]
@architectures = "spacy.TextCatBOW.v2"
@architectures = "spacy.TextCatBOW.v3"
exclusive_classes = true
length = 262144
ngram_size = 1
no_output_layer = false

10
website/meta/languages.json
View File

@ -103,6 +103,10 @@
"has_examples": true,
"models": ["fi_core_news_sm", "fi_core_news_md", "fi_core_news_lg"]
},
{
"code": "fo",
"name": "Faroese"
},
{
"code": "fr",
"name": "French",
@ -290,6 +294,12 @@
"example": "Dit is een zin.",
"has_examples": true
},
{
"code": "nn",
"name": "Norwegian Nynorsk",
"example": "Det er ein meir enn i same periode i fjor.",
"has_examples": true
},
{
"code": "pl",
"name": "Polish",

4
website/meta/site.json
View File

@ -66,6 +66,10 @@
{
"text": "Stack Overflow",
"url": "http://stackoverflow.com/questions/tagged/spacy"
},
{
"text": "Merchandise",
"url": "https://explosion.ai/merch"
}
]
},

17
website/meta/universe.json
View File

@ -4500,6 +4500,23 @@
"website": "https://nlp.unibuc.ro/people/snisioi.html"
},
"category": ["pipeline", "training", "models"]
},
{
"id": "redfield-spacy-nodes",
"title": "Redfield NLP Nodes for KNIME",
"slogan": "Makes the functionality of the spaCy library available in KNIME Analytics Platform.",
"description": "This extension provides nodes that make the functionality of the spaCy library available in the [KNIME Analytics Platform](https://www.knime.com/).",
"github": "Redfield-AB/Spacy-Nodes",
"url": "https://redfield.ai/spacy-redfield/",
"thumb": "https://raw.githubusercontent.com/Redfield-AB/Spacy-Nodes/master/resource/redfield_logo_100x100.png",
"image": "https://raw.githubusercontent.com/Redfield-AB/Spacy-Nodes/master/resource/screen1.png",
"author": "Redfield AB",
"author_links": {
"twitter": "Redfield_AB",
"github": "Redfield-AB",
"website": "https://redfield.ai"
},
"category": ["standalone"]
}
],