Skip to main content

A python module for word inflections designed for use with Spacy.

Project description


A python module for word inflections that works as a Spacy extension.

This module is designed as an extension for Spacy and will return the the inflected form of a word based on a supplied Penn Treekbank part-of-speech tag. It can also be used a stanalone module outside of Spacy.

It is based on the Automatically Generated Inflection Database (AGID). The AGID data provides a list of inflections for various word lemma.

See the scripts directory for utilities that make good examples or the tests directory for unit tests / examples.


pip3 install pyinflect

Usage as an Extension to Spacy

To use as an extension to Spacy, first import the module. This will create a new inflect method for each Spacy Token that takes in a Penn Treebank tag as its parameter. The method returns the inflected form of the token's lemma based on the supplied treekbank tag. When more than one spelling/form exists for the given tag, an optional form number parameter can be supplied, otherwise the first one is returned.

> import pyinflect
> doc = nlp('This is an example.')
> doc[3]._.inflect('NNS')

> doc[1]._.inflect('VBD', 0)

> doc[1]._.inflect('VBD', 1)

Usage Standalone

To use standalone, import the method getInflections and call it directly. getAllInflections returns all entries in the infl.csv file as a dictionary of inflected forms, where each form entry is a tuple with one or more spellings/forms for a given treebank tag. Optional parameters pos_type (which is V, A or N) and tag (the Penn treebank tag) can be used to limit the inflections returned.

> from pyinflect import getInflections
> getAllInflections('watch')
{'NN': ('watch',), 'NNS': ('watches',), 'VB': ('watch',), 'VBP': ('watch',), 'VBD': ('watched',), 'VBN': ('watched',), 'VBG': ('watching',), 'VBZ': ('watches',)}

> getAllInflections('watch', pos_type='V')
{'VB': ('watch',), 'VBP': ('watch',), 'VBD': ('watched',), 'VBN': ('watched',), 'VBG': ('watching',), 'VBZ': ('watches',)}

> getAllInflections('watch', tag='VBD')
{'VBD': ('watched',)}


If you find a bug, please report it on the Github issues list. However, there are some inflections which are ambiguous because of different spellings or because multiple forms exist for the same lemma / treebank tag. For these instances extra logic will need to be supplied to determine which one to use. Alternately the overrides.csv can be used to specify the preferred string the system returns.

See KnownIssues.txt for a list of lemma / inflections that may be problematic.

One common issues is that forms of the verb "be" are not completely specified by the treekbank tag. When the inflected form is ambiguous the first person form is returned. Setting the form_num to the Spacy inflection method allows returning the 2nd person version.

The AGID data is created by a 3rd party and not maintained here. Some lemmas are not in that data file, infl.csv, and thus can not be inflected. In some cases the AGID may not contain the best inflection of the word. For instance, lemma "people" with tag "NNS" will return "peoples" where you may want the word "people" which is also plural.


The module determines the inflection(s) returned by either a pos_type or a Penn Treebank tag. The pos_type is either 'V', A' or 'N' for 'Verb', 'Adjective'/'Adverb' or 'Noun' respectively. A list of treebank tags can be found here. Not all of these are used by pyinflect. The following is a list of the various types and tags used...

pos_type = 'A'
* JJ      Adjective
* JJR     Adjective, comparative
* JJS     Adjective, superlative
* RB      Adverb
* RBR     Adverb, comparative
* RBS     Adverb, superlative

pos_type = 'N'
* NN      Noun, singular or mass
* NNS     Noun, plural

pos_type = 'V'
* VB      Verb, base form
* VBD     Verb, past tense
* VBG     Verb, gerund or present participle
* VBN     Verb, past participle
* VBP     Verb, non-3rd person singular present
* VBZ     Verb, 3rd person singular present
* MD      Modal

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

pyinflect-0.3.0.tar.gz (7.3 kB view hashes)

Uploaded source

Built Distribution

pyinflect-0.3.0-py3-none-any.whl (700.5 kB view hashes)

Uploaded py3

Supported by

AWS AWS Cloud computing Datadog Datadog Monitoring Facebook / Instagram Facebook / Instagram PSF Sponsor Fastly Fastly CDN Google Google Object Storage and Download Analytics Huawei Huawei PSF Sponsor Microsoft Microsoft PSF Sponsor NVIDIA NVIDIA PSF Sponsor Pingdom Pingdom Monitoring Salesforce Salesforce PSF Sponsor Sentry Sentry Error logging StatusPage StatusPage Status page