Is there a way to regex graphemes? #104
Replies: 2 comments
-
Related discussion here: #62 (comment) |
Beta Was this translation helpful? Give feedback.
-
Sorry, I missed the notification for this; this would be a deviation from the PLS standard, so it's something that needs to be done carefully. I see the use case, since there are too many permutations of words and punctuation to work within the existing standard on this. On the other hand, allowing regex in our PLS will make our version of PLS incompatible with standard PLS engines, which I feel matters in terms of community stewardship (if someone tries using that lexicon in another program, it will break unexpectedly).
Given that we already have our own PLS engine that constructs well-formed SSML, supporting this for ourselves shouldn't be a problem. Some possible solutions:
Converting this into an issue. |
Beta Was this translation helpful? Give feedback.
-
I was wondering if there was a way to use a regex to only replace a grapheme when the regex applies? So I could make the word
hm
get replaced while not messing with the wordhmm
orRhythm
. If some sort of Regex is currently possible I could just regex\s*[hH]m\W
and then replace pronunciation ofhm
only after the regex matches.The process I was hoping for would be:
If this is either already possible, or able to be implemented, it'll be very helpful for short words that are pronounced weirdly, and avoid those short graphemes interacting with longer words that are not included in the lexicon.
Example, a grapheme for "tis", would interact with satisfy. and would make is say " Sa tis F Y"
ffxiv_dx11_DPTelr2IdT.mp4
(May need to unmute)
It would also help with Characters and races with a
's
contained within their names, that are read normally without needing correcting(thus not in the lexicon) getting broken in the same way with the grapheme's
.I could see possible issues with putting regex into the .pls file if a voice backend isn't expecting those contents and needing a separate file for regexing. If it's already possible though, it may not be a problem, and may already be built into AWS and the system backend. I'm not sure, let me know if you know anything!
Beta Was this translation helpful? Give feedback.
All reactions