Wn is a Python library for using wordnets. For example:

>>> import wn
>>> wn.download('ewn:2020')    # Install the English Wordnet 2020 (only once)
Download complete (13643357 bytes)
Checking /tmp/tmpgspkay6m.xml
Reading /tmp/tmpgspkay6m.xml
Building: [###############################] (1337590/1337590)
>>> ss = wn.synsets('win')[0]  # Get the first synset for 'win'
>>> ss.definition()            # Get the synset's definition
'be the winner in a contest or competition; be victorious'

Unlike previous implementations, Wn uses a SQLite database to store wordnet data, which can make it much faster: Wn is 5x faster than the NLTK to list all English synsets, and almost 20x faster if you include the startup time. Some operations, particularly path operations that require multiple SQL queries, may be slower.

Wn is also multilingual from the start. English is not the default. Instead, all wordnets are searched unless one (or more) are specified:

>>> from nltk.corpus import wordnet as nltk_wn
>>> nltk_wn.synsets('chat')                 # only English
>>> nltk_wn.synsets('chat', lang='fra')     # only French
>>> import wn
>>> wn.synsets('chat')                      # all installed wordnets
>>> wn.synsets('chat', lgcode='en')         # limit to one language
>>> wn.synsets('chat', lexicon='ewn:2020')  # limit to one wordnet

Currently Wn is only available via this repository:

pip install git+https://github.com/goodmami/wn.git

You can also clone the repository and use flit to install it:

git clone https://github.com/goodmami/wn.git
cd wn
pip install flit  # if you don't have it
flit install

The documentation is hosted here, although it's not very useful yet.

The following wordnets are indexed by Wn and ready to be installed:

(more coming soon)

Some operations keep a compatible API with the NLTK's wordnet module, but most will need some translation.

(this table is incomplete)