Alternative Pronunciations within the dictionary

[Caitlin Halfacre]

Can anyone tell me how MFA works with alternative pronunciations in the dictionary? I noticed that for some items in the dictionary there are multiple pronunciations (e.g. 'present' shown below) so I wanted to try expanding this. Since my speakers are from England they generally don't pronounce coda /r/ (except in linking position) so using some regex search and replace commands I made duplicates of words with non rhotic pronunciations (e.g. first shown below). However, when I run the aligner, the alternative (non-rhotic) pronunciation is never used. Is this because the inbuilt acoustic model is American English and so it automatically looks for presence of /r/ or because the aligner doesn't actually have a way of handling more than one entry in the pronouncing dictionary so just ignores it? I'd like to try alternative pronunciations for other things, e.g. B AE1 TH vs B AA1 TH, or t,d, deletion (as Bailey 2016 uses to show other aligners can be used to find variation) so would like to know how MFA handles these double entries.

PRESENT  P ER0 Z EH1 N T
PRESENT  P R EH1 Z AH0 N T
PRESENT  P R IY0 Z EH1 N T
PRESENT' P R IH0 Z EH1 N

FIRST  F AH1 S T 
FIRST  F ER1 S T

[lipan...@gmail.com]

I don't exactly have the answer to your question, but I am currently working on editing the dictionary in the way you describe, and I am able to get it to recognize (with limited success...) pronunciation variants in general American English.

[Caitlin Halfacre]

You've managed to have multiple variants of one word in there? How have you formatted it? Just on separate lines like above?