I took a look at the idea mentioned by Jason and I think it does not fully meet my requirement. The discussion associated with the idea mentions this at the end of September 2022 under the term "lexicon support" and, in the end, it really seems that the only way forward is the one indicated by Vaun: probably the lexicon needs to be managed in a datatable to be accessed with the word to be transformed, retrieving its SSML "transcript".
However, I am concerned about the delay induced by this approach because, in theory, I would have to do a search in the datatable for each word of the phrase I need to play through the TTS (consider that the text to be played comes from an external source and is not known a priori).
This is probably not the best thing to do (surely the TTS engine contains within it this kind of text substitution in a much more efficient way).
------------------------------
Giuliano Ferri
Indra Italia
------------------------------
Original Message:
Sent: 03-15-2025 02:02
From: Vaun McCarthy
Subject: TTS Dictionary
Just a warning, while this says upcoming development, that could still mean many months before something usable comes to the forefront and gets released.
I don't know if this would work, but you could build your own dictionary up if the terms list won't be huge. You could put these into a datatable, and have one of the fields point at an Architect prompt that you have generated ahead of time using the TTS console from your provider (e.g. Amazon Polly) using whatever SSML tags you want. If your flow doesn't find an entry in the datatable, it'll just play TTS as it currently does. If it finds an entry in there it'll play the prompt you'd setup ahead of time for that particular "dictionary" entry.
Possibly when the idea Jason linked to gets released, you might be able to re-use this same datatable to include a field for the text with SSML tags you want to use.
------------------------------
Vaun McCarthy
Original Message:
Sent: 03-14-2025 17:27
From: Jason Kleitz
Subject: TTS Dictionary
Hello Giuliano,
I don't think that we currently have a feature like this available yet, but I do see that this Idea in the Product Ideas Lab is currently listed as Upcoming Development.
If you click the VOTE button on that page and leave a comment, it should notify you via email as this Idea progresses through development.
------------------------------
Jason Kleitz
Online Community Manager/Moderator
Original Message:
Sent: 03-14-2025 07:00
From: Giuliano Ferri
Subject: TTS Dictionary
Hello Community,
do you know if there is a way to define a dictionary list to be used when TTS need to synthesize a variable text, for example coming from an external source. We know that SSML is available but you could use it when you're writing a text on your own (and you can say to TTS how pronounce specific terms). But this approach is not suitable when a variable (and potentially unknown) text is received.
Ideally, there should be somewhere a list of terms with the corresponding pronunciation. In specific TTS products you can specify the list of phonemes to pronounce but it could be enough, in this case, adjust the pronunciation using normal characters.
Any ideas?
#ConversationalAI(Bots,AgentAssist,etc.)
#Unsure/Other
------------------------------
Giuliano Ferri
Director
------------------------------