Although the main focus in on Grapheme-to-Phoneme conversion, the approach that Phonetisaurus employs is symmetric with regard to the input and output symbols, thus it is relatively trivial to also construct Phoneme-to-Grapheme models using the same approach with albeit swapped inputs and outputs.
== Forward P2G ==
A Phoneme-to-Grapheme model can be generated from the same input dictionary by simply flipping the '--swap' flag, which will swap the positions of the pronunciations and words in the training data.
For the moment the best bet would be to use the actual older version. These are still available in the defunct downloads section in googlecode. We can see about adding them back in a bit later.
I believe old versions of Phonetisaurus had a
--swap
flag that could be used for P2G. Is this still possible?http://wiki.phonetisaurus.googlecode.com/hg-history/01f8ea06f6591cf74d8ef4fd156cd27728ec9bd5/QuickStartExamples.wiki