explosion/spaCy
1
1
Fork 0
mirror of https://github.com/explosion/spaCy.git synced 2024-12-25 17:36:30 +03:00
Code Issues Packages Projects Releases Wiki Activity

* Work on intro copy

Browse Source
This commit is contained in:
Matthew Honnibal 2014-11-03 00:13:19 +11:00
parent fa91506073
commit f1c3e17c80
1 changed files with 19 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

46
docs/source/index.rst
View File

@ -6,36 +6,28 @@
spaCy NLP Tokenizer and Lexicon
================================
spaCy is an industrial-strength multi-language tokenizer, bristling with features
you never knew you wanted. You do want these features though --- your current
tokenizer has been doing it wrong.
Where other tokenizers give you a list of strings, spaCy gives you references
to rich lexical types, for easy, excellent and efficient feature extraction.
spaCy is a library for industrial strength NLP in Python and Cython. Its core
values are efficiency, accuracy and minimalism.
* **Easy**: Tokenizer returns a sequence of rich lexical types, with features
pre-computed:
* Efficiency: spaCy is
>>> from spacy.en import EN
>>> for w in EN.tokenize(string):
... print w.sic, w.shape, w.cluster, w.oft_title, w.can_verb
Check out the tutorial and API docs.
* **Excellent**: Distributional and orthographic features are crucial to robust
NLP. Without them, models can only learn from tiny annotated training
corpora. Read more.
* **Efficient**: spaCy serves you rich lexical objects faster than most
tokenizers can give you a list of strings.
+--------+-------+--------------+--------------+
| System | Time | Words/second | Speed Factor |
+--------+-------+--------------+--------------+
| NLTK | 6m4s | 89,000 | 1.00 |
+--------+-------+--------------+--------------+
| spaCy | 9.5s | 3,093,000 | 38.30 |
+--------+-------+--------------+--------------+
It does not attempt to be comprehensive,
or to provide lavish syntactic sugar. This isn't a library that covers 43 known
algorithms to do X. You get 1 --- the best one --- with a simple, low-level interface.
For commercial users, the code is free but the data isn't. For researchers, both
are free and always will be.
Comparison
----------
+-------------+-------------+---+-----------+--------------+
| POS taggers | Speed (w/s) | % Acc. (news) | % Acc. (web) |
+-------------+-------------+---------------+--------------+
| spaCy | | | |
+-------------+-------------+---------------+--------------+
| Stanford | 16,000 | | |
+-------------+-------------+---------------+--------------+
| NLTK | | | |
+-------------+-------------+---------------+--------------+
.. toctree::