spaCy: Industrial-strength NLP
spaCy is a library for advanced natural language processing in Python and
Cython. spaCy is built on the very latest research, but it isn't researchware.
It was designed from day one to be used in real products. spaCy currently supports
English, German, French and Spanish, as well as tokenization for Italian,
Portuguese, Dutch, Swedish, Finnish, Norwegian, Hungarian, Bengali, Hebrew,
Chinese and Japanese. It's commercial open-source software, released under the
MIT license.
βοΈ Test spaCy v2.0.0 alpha and the new models! Read the release notes here.
π« Version 1.8 out now! Read the release notes here.
π¬ Where to ask questions
Non-destructive tokenization
Syntax-driven sentence segmentation
Pre-trained word vectors
Part-of-speech tagging
Named entity recognition
Labelled dependency parsing
Convenient string-to-int mapping
Export to numpy data arrays
GIL-free multi-threading
Efficient binary serialization
Easy deep learning integration
Statistical models for English , German , French and Spanish
State-of-the-art speed
Robust, rigorously evaluated accuracy
See facts, figures and benchmarks .
Fastest in the world: <50ms per document. No faster system has ever been
announced.
Accuracy within 1% of the current state of the art on all tasks performed
(parsing, named entity recognition, part-of-speech tagging). The only more
accurate systems are an order of magnitude slower or more.
Operating system
macOS / OS X, Linux, Windows (Cygwin, MinGW, Visual Studio)
Python version
CPython 2.6, 2.7, 3.3+. Only 64 bit.
Package managers
pip (source packages only), conda (via conda-forge)
Installation requires a working build environment. See notes on Ubuntu,
macOS/OS X and Windows for details.
Using pip, spaCy releases are currently only available as source packages.
When using pip it is generally recommended to install packages in a virtualenv
to avoid modifying system state:
virtualenv .env
source .env/bin/activate
pip install spacy
Thanks to our great community, we've finally re-added conda support. You can now
install spaCy via conda-forge:
conda config --add channels conda-forge
conda install spacy
For the feedstock including the build recipe and configuration,
check out this repository .
Improvements and pull requests to the recipe and setup are always appreciated.
As of v1.7.0, models for spaCy can be installed as Python packages .
This means that they're a component of your application, just like any
other module. They're versioned and can be defined as a dependency in your
requirements.txt. Models can be installed from a download URL or
a local directory, manually or via pip. Their data can be located anywhere on
your file system. To make a model available to spaCy, all you need to do is
create a "shortcut link", an internal alias that tells spaCy where to find the
data files for a specific model name.
# out-of-the-box: download best-matching default model
python -m spacy download en
# download best-matching version of specific model for your spaCy installation
python -m spacy download en_core_web_md
# pip install .tar.gz archive from path or URL
pip install /Users/you/en_core_web_md-1.2.0.tar.gz
pip install https://github.com/explosion/spacy-models/releases/download/en_core_web_md-1.2.0/en_core_web_md-1.2.0.tar.gz
# set up shortcut link to load installed package as "en_default"
python -m spacy link en_core_web_md en_default
# set up shortcut link to load local model as "my_amazing_model"
python -m spacy link /Users/you/data my_amazing_model
To load a model, use spacy.load() with the model's shortcut link:
import spacy
nlp = spacy .load ('en_default' )
doc = nlp (u'This is a sentence.' )
If you've installed a model via pip, you can also import it directly and
then call its load() method with no arguments. This should also work for
older models in previous versions of spaCy.
import spacy
import en_core_web_md
nlp = en_core_web_md .load ()
doc = nlp (u'This is a sentence.' )
π For more info and examples, check out the models documentation .
Support for older versions
If you're using an older version (v1.6.0 or below), you can still download and
install the old models from within spaCy using python -m spacy.en.download all
or python -m spacy.de.download all. The .tar.gz archives are also
attached to the v1.6.0 release .
To download and install the models manually, unpack the archive, drop the
contained directory into spacy/data and load the model via spacy.load('en')
or spacy.load('de').
The other way to install spaCy is to clone its
GitHub repository and build it from
source. That is the common way if you want to make changes to the code base.
You'll need to make sure that you have a development enviroment consisting of a
Python distribution including header files, a compiler,
pip , virtualenv
and git installed. The compiler part is the trickiest.
How to do that depends on your system. See notes on Ubuntu, OS X and Windows for
details.
# make sure you are using recent pip/virtualenv versions
python -m pip install -U pip virtualenv
git clone https://github.com/explosion/spaCy
cd spaCy
virtualenv .env
source .env/bin/activate
pip install -r requirements.txt
pip install -e .
Compared to regular install via pip requirements.txt
additionally installs developer dependencies such as Cython.
Instead of the above verbose commands, you can also use the following
Fabric commands:
fab env
Create virtualenv and delete previous one, if it exists.
fab make
Compile the source.
fab clean
Remove compiled objects, including the generated C++.
fab test
Run basic tests, aborting after first failure.
All commands assume that your virtualenv is located in a directory .env.
If you're using a different directory, you can change it via the environment
variable VENV_DIR, for example:
VENV_DIR=" .custom-env" fab clean make
Install system-level dependencies via apt-get:
sudo apt-get install build-essential python-dev git
Install a recent version of XCode ,
including the so-called "Command Line Tools". macOS and OS X ship with Python
and git preinstalled.
Install a version of Visual Studio Express
or higher that matches the version that was used to compile your Python
interpreter. For official distributions these are VS 2008 (Python 2.7),
VS 2010 (Python 3.4) and VS 2015 (Python 3.5).
spaCy comes with an extensive test suite . First, find out where
spaCy is installed:
python -c " import os; import spacy; print(os.path.dirname(spacy.__file__))"
Then run pytest on that directory. The flags --vectors, --slow
and --model are optional and enable additional tests:
# make sure you are using recent pytest version
python -m pip install -U pytest
python -m pytest < spacy-directory> --vectors --models --slow
Version
Date
Description
v1.8.2
2017-04-26
French model and small improvements
v1.8.1
2017-04-23
Saving, loading and training bug fixes
v1.8.0
2017-04-16
Better NER training, saving and loading
v1.7.5
2017-04-07
Bug fixes and new CLI commands
v1.7.3
2017-03-26
Alpha support for Hebrew, new CLI commands and bug fixes
v1.7.2
2017-03-20
Small fixes to beam parser and model linking
v1.7.1
2017-03-19
Fix data download for system installation
v1.7.0
2017-03-18
New 50 MB model, CLI, better downloads and lots of bug fixes
v1.6.0
2017-01-16
Improvements to tokenizer and tests
v1.5.0
2016-12-27
Alpha support for Swedish and Hungarian
v1.4.0
2016-12-18
Improved language data and alpha Dutch support
v1.3.0
2016-12-03
Improve API consistency
v1.2.0
2016-11-04
Alpha tokenizers for Chinese, French, Spanish, Italian and Portuguese
v1.1.0
2016-10-23
Bug fixes and adjustments
v1.0.0
2016-10-18
Support for deep learning workflows and entity-aware rule matcher
v0.101.0
2016-05-10
Fixed German model
v0.100.7
2016-05-05
German support
v0.100.6
2016-03-08
Add support for GloVe vectors
v0.100.5
2016-02-07
Fix incorrect use of header file
v0.100.4
2016-02-07
Fix OSX problem introduced in 0.100.3
v0.100.3
2016-02-06
Multi-threading, faster loading and bugfixes
v0.100.2
2016-01-21
Fix data version lock
v0.100.1
2016-01-21
Fix install for OSX
v0.100
2016-01-19
Revise setup.py, better model downloads, bug fixes
v0.99
2015-11-08
Improve span merging, internal refactoring
v0.98
2015-11-03
Smaller package, bug fixes
v0.97
2015-10-23
Load the StringStore from a json list, instead of a text file
v0.96
2015-10-19
Hotfix to .merge method
v0.95
2015-10-18
Bug fixes
v0.94
2015-10-09
Fix memory and parse errors
v0.93
2015-09-22
Bug fixes to word vectors