Machine Transliterating into Conlangs

I’ve been writing a tool kit of utilities for toki pona with an eye to making these potentially general tools for any conlang. One of them is a transliterator that transliterates any word into toki pona phonotactics, e.g. Clifford becomes Kipo.

One problem is that English isn’t spelled phonetically, so to get the transliterator to work, I’m adding lines of code that say, “tion” is “shun” This doesn’t work very well, as there are a heck of a lot of oddities in the spelling system. IPA to the rescue right? Well, first I need a publicly available dictionary of IPA pronunciations.

ARPAbet is a phonetic alphabet that for the most part can be transliterated into IPA- but AFAIK, it only covers English phonemes.

And CMU has released a pronunciation dictionary that uses ARPAbet

So now if I can map ARPAbet to toki pona, I can focus on transliteration issues and not so much on how to cope with English spelling.

By the way, if you are interested in alternate ASCII representions of IPA, there is


Taste may vary, but I think of the three, ARPAbet is the least ugly. It would have to be extended to use it for general conlang use, but its worth knowing about.

This entry was posted in conlang design, machine assisted conlanging. Bookmark the permalink.

Comments are closed.