Skip to content
Rule-based pronunciation for English
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Type Name Latest commit message Commit time
Failed to load latest commit information.

Elovitz: rule-based English pronunciation

This is code which implements a rule-based system for determining the pronunciation of English words. It's based on the algorithm described in this 1976 paper by Elovitz et al.

Example usage:

>>> import elovitz
>>> elovitz.pronounce('unique')
['y', 'uw', 'n', 'iy', 'k']
>>> elovitz.ipa('unique')

Support, licensing, ongoing development

This is code I wrote for a specific project, which I have now finished.

The code is available under the MIT license, so you can fork it, improve it, learn from it, build upon it. However, I have no interest in maintaining it as an ongoing open source project, nor in providing support for it. Pull requests will probably be either ignored or closed.

If you do make something interesting with this code, please do still let me know! I'm sorry that I can't provide any support, but I am still genuinely interested in seeing creative applications of the code.

You can’t perform that action at this time.