See readme_install.org for installation of Pynini on Ubuntu 14.04 using vagrant.
- first examples
- singularization
- edit-transducer tutorial
- edit transducer class
- Levenshtein examples
- Levenshtein Automaton for american english words
- Levenshtein Automaton for mw headwords
The mw headwords example concluded that it takes about 5 sec to get closest spellings for a (misspelled) word, in this lexicon of about 200,000 words. This is too long to be very useful when checking more than a few words. Perhaps there is some additional optimization of the FST.