I would like to take snippets of text and convert them, programmatically, to be more phonetic than traditional English spellings. The purpose of this is for converting a bunch of text I’m working with to a version that can be read by common TTS tools in a more natural way for a scientific application.
Many of the terms and acronyms are common, but there are a variety that are not. I am hoping to find a script or resource of some kind that already has much of this effort done (realizing I will have to do a bit of customization as per these terms), however so far I’ve found nothing that goes down this path.
I HAVE found solutions that are truly phonetic and are for linguistic applications, however these are not desirable as they are not readable by standard off the shelf TTS solutions that I’ve used.
Any ideas of a starting point for this situation? Any examples or even libs that make this easier to chew would be fine. Or am I bound to sit down and grind out a solution entirely of my own design?
Here is an idea that should work :
Use a cunning, “linguist”-grade phoneticizer.
Use a standard, small table of phonetic to English syllable ( in the English accent of your choosing ), mappings.
Run the resulting “phonetically spelled” words through a standard, OTS TTS product .
So, for example, you can have :
Fish and chips —-step 1—> phonetic linguist code —step 2. (new Zealand accent)—> fush und chupsh —step 3. —> audio pleasure
Hope this assists you!