explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2024-12-24 17:06:29 +03:00
Code Issues Packages Projects Releases Wiki Activity

Rename "English" section to "Language"

Browse Source
This commit is contained in:
Ines Montani 2016-10-21 00:58:24 +02:00
parent e16e78a737
commit f8322a69e7
3 changed files with 101 additions and 117 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

214
website/docs/_api-english.jade → website/docs/_api-language.jade
View File

@ -1,150 +1,134 @@
//- ----------------------------------
//- 💫 DOCS > API > ENGLISH
//- 💫 DOCS > API > LANGUAGE
//- ----------------------------------
+section("english")
+h(2, "english", "https://github.com/" + SOCIAL.github + "/spaCy/blob/master/spacy/language.py")
| #[+tag class] English(Language)
+section("language")
+h(2, "language", "https://github.com/" + SOCIAL.github + "/spaCy/blob/master/spacy/language.py")
| #[+tag class] Language
p.
The English analysis pipeline. Usually you"ll load this once per process,
and pass the instance around your program.
A pipeline that transforms text strings into annotated spaCy Doc objects. Usually you'll load the Language pipeline once and pass the instance around your program.
+code("python", "Overview").
class Language:
lang = None
def __init__(self, data_dir=None, tokenizer=None, tagger=None, parser=None, entity=None, matcher=None):
return self
Defaults = BaseDefaults
def __call__(self, text, tag=True, parse=True, entity=True):
return Doc()
def __init__(self, path=True, **overrides):
self.vocab = Vocab()
self.tokenizer = Tokenizer()
self.tagger = Tagger()
self.parser = DependencyParser()
self.entity = EntityRecognizer()
self.make_doc = lambda text: Doc()
self.pipeline = [self.tagger, self.parser, self.entity]
def pipe(self, texts_iterator, batch_size=1000, n_threads=2):
yield Doc()
def __call__(self, text, **toggle):
doc = self.make_doc(text)
for proc in self.pipeline:
if toggle.get(process.name, True):
process(doc)
return doc
def end_training(self, data_dir=None):
def pipe(self, texts_iterator, batch_size=1000, n_threads=2, **toggle):
docs = (self.make_doc(text) for text in texts_iterator)
for process in self.pipeline:
if toggle.get(process.name, True):
docs = process.pipe(docs, batch_size=batch_size, n_threads=n_threads)
for doc in self.docs:
yield doc
def end_training(self, path=None):
return None
class English(Language):
lang = "en"
class English(Language):
class Defaults(BaseDefaults):
pass
class German(Language):
lang = "de"
class German(Language):
class Defaults(BaseDefaults):
pass
+section("english-init")
+h(3, "english-init")
| #[+tag method] English.__init__
| #[+tag method] Language.__init__
p
| Load the pipeline. Each component can be passed
| as an argument, or left as #[code None], in which case it will be loaded
| from a classmethod, named e.g. #[code default_vocab()].
| Load the pipeline. You can disable components by passing None as a value,
| e.g. pass parser=None, vectors=None to save memory if you're not using
| those components. You can also pass an object as the value.
| Pass a function create_pipeline to use a custom pipeline --- see
| the custom pipeline tutorial.
+aside("Efficiency").
Loading takes 10-20 seconds, and the instance consumes 2 to 3
gigabytes of memory. Intended use is for one instance to be
created for each language per process, but you can create more
if you"re doing something unusual. You may wish to make the
if you're doing something unusual. You may wish to make the
instance a global variable or "singleton".
+table(["Example", "Description"])
+row
+cell #[code.lang-python nlp = English()]
+cell Load everything, from default package
+cell #[code nlp = English()]
+cell Load everything, from default path.
+row
+cell #[code.lang-python nlp = English(data_dir='my_data')]
+cell Load everything, from specified dir
+cell #[code nlp = English(path='my_data')]
+cell Load everything, from specified path
+row
+cell #[code.lang-python nlp = English(parser=False)]
+cell Load everything except the parser.
+cell #[code nlp = English(path=path_obj)]
+cell Load everything, from an object that follows the #[code pathlib.Path] protocol.
+row
+cell #[code.lang-python nlp = English(parser=False, tagger=False)]
+cell Load everything except the parser and tagger.
+cell #[code nlp = English(parser=False, vectors=False)]
+cell Load everything except the parser and the word vectors.
+row
+cell #[code.lang-python nlp = English(parser=MyParser())]
+cell Supply your own parser
+cell #[code nlp = English(parser=my_parser)]
+cell Load everything, and use a custom parser.
+row
+cell #[code nlp = English(create_pipeline=my_pipeline)]
+cell Load everything, and use a custom pipeline.
+code("python", "Definition").
def __init__(self, data_dir=None, tokenizer=None, tagger=None, parser=None, entity=None, matcher=None):
return self
def __init__(self, path=True, **overrides):
D = self.Defaults
self.vocab = Vocab(path=path, parent=self, **D.vocab) \
if 'vocab' not in overrides \
else overrides['vocab']
self.tokenizer = Tokenizer(self.vocab, path=path, **D.tokenizer) \
if 'tokenizer' not in overrides \
else overrides['tokenizer']
self.tagger = Tagger(self.vocab, path=path, **D.tagger) \
if 'tagger' not in overrides \
else overrides['tagger']
self.parser = DependencyParser(self.vocab, path=path, **D.parser) \
if 'parser' not in overrides \
else overrides['parser']
self.entity = EntityRecognizer(self.vocab, path=path, **D.entity) \
if 'entity' not in overrides \
else overrides['entity']
self.matcher = Matcher(self.vocab, path=path, **D.matcher) \
if 'matcher' not in overrides \
else overrides['matcher']
+table(["Arg", "Type", "Description"])
+row
+cell data_dir
+cell str
+cell.
The data directory. If None, value is obtained via the
#[code default_data_dir()] method.
if 'make_doc' in overrides:
self.make_doc = overrides['make_doc']
elif 'create_make_doc' in overrides:
self.make_doc = overrides['create_make_doc'](self)
else:
self.make_doc = lambda text: self.tokenizer(text)
if 'pipeline' in overrides:
self.pipeline = overrides['pipeline']
elif 'create_pipeline' in overrides:
self.pipeline = overrides['create_pipeline'](self)
else:
self.pipeline = [self.tagger, self.parser, self.matcher, self.entity]
+row
+cell vocab
+cell #[code Vocab]
+cell.
The vocab object, which should be an instance of class
#[code spacy.vocab.Vocab]. If #[code None], the object is
obtained from the #[code default_vocab()] class method. The
vocab object manages all of the language specific rules and
definitions, maintains the cache of lexical types, and manages
the word vectors. Because the vocab owns this important data,
most objects hold a reference to the vocab.
+row
+cell tokenizer
+cell #[code Tokenizer]
+cell.
The tokenizer, which should be a callable that accepts a
unicode string, and returns a #[code Doc] object. If set to
#[code None], the default tokenizer is constructed from the
#[code default_tokenizer()] method.
+row
+cell tagger
+cell #[code Tagger]
+cell.
The part-of-speech tagger, which should be a callable that
accepts a #[code Doc] object, and sets the part-of-speech
tags in-place. If set to None, the default tagger is constructed
from the #[code default_tagger()] method.
+row
+cell parser
+cell #[code Parser]
+cell.
The dependency parser, which should be a callable that accepts
a #[code Doc] object, and sets the sentence boundaries,
syntactic heads and dependency labels in-place.
If set to #[code None], the default parser is
constructed from the #[code default_parser()] method. To disable
the parser and prevent it from being loaded, pass #[code parser=False].
+row
+cell entity
+cell #[code Parser]
+cell.
The named entity recognizer, which should be a callable that
accepts a #[code Doc] object, and sets the named entity annotations
in-place. If set to None, the default entity recognizer is
constructed from the #[code default_entity()] method. To disable
the entity recognizer and prevent it from being loaded, pass
#[code entity=False].
+row
+cell matcher
+cell #[code Matcher]
+cell.
The pattern matcher, which should be a callable that accepts
a #[code Doc] object, and sets named entity annotations in-place
using token-based rules. If set
to None, the default matcher is constructed from the
#[code default_matcher()] method.
+section("english-call")
+h(3, "english-call")
| #[+tag method] English.__call__
+section("language-call")
+h(3, "language-call")
| #[+tag method] Language.__call__
p
| The main entry point to spaCy. Takes raw unicode text, and returns
@ -152,30 +136,30 @@
| and #[code Span] objects.
+aside("Efficiency").
spaCy"s algorithms are all linear-time, so you can supply
spaCy's algorithms are all linear-time, so you can supply
documents of arbitrary length, e.g. whole novels.
+table(["Example", "Description"], "code")
+row
+cell #[code.lang-python doc = nlp(u'Some text.')]
+cell #[ doc = nlp(u'Some text.')]
+cell Apply the full pipeline.
+row
+cell #[code.lang-python doc = nlp(u'Some text.', parse=False)]
+cell #[ doc = nlp(u'Some text.', parse=False)]
+cell Applies tagger and entity, not parser
+row
+cell #[code.lang-python doc = nlp(u'Some text.', entity=False)]
+cell #[ doc = nlp(u'Some text.', entity=False)]
+cell Applies tagger and parser, not entity.
+row
+cell #[code.lang-python doc = nlp(u'Some text.', tag=False)]
+cell #[ doc = nlp(u'Some text.', tag=False)]
+cell Does not apply tagger, entity or parser
+row
+cell #[code.lang-python doc = nlp(u'')]
+cell #[ doc = nlp(u'')]
+cell Zero-length tokens, not an error
+row
+cell #[code.lang-python doc = nlp(b'Some text')]
+cell #[ doc = nlp(b'Some text')]
+cell Error: need unicode
+row
+cell #[code.lang-python doc = nlp(b'Some text'.decode('utf8'))]
+cell #[ doc = nlp(b'Some text'.decode('utf8'))]
+cell Decode bytes into unicode first.
+code("python", "Definition").

2
website/docs/_data.json
View File

@ -8,7 +8,7 @@
["Usage Examples", "#examples", "examples"]
],
"API": [
["English", "#english", "english"],
["Language", "#language", "language"],
["Doc", "#doc", "doc"],
["Token", "#token", "token"],
["Span", "#span", "span"],

2
website/docs/index.jade
View File

@ -13,7 +13,7 @@ include _quickstart-examples
+h(2, "api") API
include _api-english
include _api-language
include _api-doc
include _api-token
include _api-span