Re: Automatic IPA transcription
Christian Pietsch <chr.pietsch <at> googlemail.com>
2012-06-20 19:17:57 GMT
I assume you have English text, not speech. Then what you need is a
grapheme-to-phoneme (G2P) converter. You will find them as components
of text-to-speech (TTS) systems. For English text, you could use
eSpeak or Festival, both of which are easily obtainable, e.g. as
Debian or Ubuntu Linux packages. Here is something I tried:
$ echo 'Will you pronounce this correctly?' | espeak -v en -x -q
--> wIl ju: pr <at> n'aUns DIs k <at> r'Ektli
The output you can see here is what eSpeak calls “phoneme mnemonics”,
but I guess it is X-SAMPA which is an ASCII representation of IPA. For
a mapping table and code in several programming languages, including
Python, see Henrik Theiling's IPA site <http://www.theiling.de/ipa/>.
Using his cxs.py module and CXS.def lookup table, I get this result:
--> wɪl juː prənˈaʊns ðɪs kərˈɛktli
Looks OK to me.
Instead of using parts of a full TTS system, you can also use
stand-alone G2P tools such as Sequitur G2P or Phonetisaurus, but you
might have to train them first.
Hope this helps,
On Tue, Jun 19, 2012 at 02:23:30PM -0400, Sam Raker wrote:
> I was wondering if anyone has found a good (OSX/*NIX-compatible)
> program for automatic transcription (of English) to IPA. There are a