Wanted 🕵️‍♀ : Translator for SC-Voice interface

[Gabriel_L]

AN10.2 has it.

[sabbamitta]

@Gabriel_L

AN10.2 has it.

an10.2:1.0.6; Ricardo; êxtase: see below

[Gabriel_L]

It should read aecstase or ɛkstɑz

https://michaelis.uol.com.br/palavra/MYPoE/extase/

[Gabriel_L]

Happy to help! Could you kindly isolate and share the file to be translated as we did previously?
Also, could you or @karl_lew let me know once it is released ? I will share the news in other forums and see if I can get others to help checking Ricardo’s pronunciation.

[karl_lew]

@Gabriel_L

The only quirk was how it pronounces êxtase

Here are some possibilities:

1. original

1. <phoneme alphabet="ipa" ph="'o"/> <phoneme alphabet="ipa" ph="'ɛχ.tase"/>

1. <phoneme alphabet="ipa" ph="'ɛχ.tase"/>

A subtlety here is that the Portuguese conjunction “o” closely binds the joined words, which works well in a colloquial phrasing where “o” is used for emphasis by reiterating a concept using synonyms. In the suttas, the “or” is enumerative with distinct and precise alternatives. Because of this, I experimented with a more emphatic “o” by making its very own phoneme in the second example. To get the longer start to êxtase I added the primary emphasis mark to the IPA in the second and third examples.

Here is the updated excerpt from AN10.2/pt/beisert:

For “arahant”, I split up the syllables, but had difficulty avoiding the Portuguese preference for turning an ending “t” into a “tch”. The only way I could get an actual “t” was to add an “e”. Is <phoneme alphabet="ipa" ph="'aɾ.han.te"/> acceptable?

Note that the changes made are universal, not contextual. This means that we’ll need to amend these pronunciations carefully, mindful of all contexts.

[karl_lew]

@scatterbrain

the SSML phoneme tag,

That is indeed how Voice pokes and prods the AI voices into something resembling what we want. If you are inclined to delve into correcting the pronunciations for the translators using Polly SSML, your help would be much appreciated. It takes a discerning ear and a good deal of patience with trial and error. The IPA is not always honored phoneme by phoneme. The AI tends to introduce contextual bias that will cause the same phoneme to be pronounced differently according to context. Would you be interested in helping with the SSML?

[sabbamitta]

@Gabriel_L

Also, could you or @karl_lew let me know once it is released ?

Updated staging server for you to test. If that’s fine we can update on production so that it’s easier for you to share on other forums.

[Gabriel_L]

Thanks for working on it so promptly @karl_lew!

Unfortunately both above are not yet sounding right.

Êxtase should sound like this

Arahant should sound like this

[sabbamitta]

@sabbamitta

If that’s fine

@Gabriel_L

Unfortunately both above are not yet sounding right.

Obviously it’s rather not fine.

We’ll try again. That’s just the normal business with getting pronunciation right. It is certainly more difficult for Karl if it’s a language he doesn’t speak, so it is good if you can share examples how it should sound!

Is “Arahant” also a word that is used in your translations? If it’s more than just one or two issues, we’ll probably defer a more thorough Portuguese pronunciation check to later; this will also give you more time to try out and listen—the more ears the better. We can collect all the instances and work on them at a later point.

[Gabriel_L]

@sabbamitta

Is “Arahant” also a word that is used in your translations? If it’s more than just one or two issues, we’ll probably defer a more thorough Portuguese pronunciation check to later; this will also give you more time to try out and listen—the more ears the better.

Yes, arahant will be a term used in the translations some times.

To the point of logging and improving the pronunciations, is it the case we start a new topic just to capture any further pronunciation issues I or anyone else finds with Ricardo’s reading of the suttas?

The reason I ask is that it would be nice to point to a specific topic/thread when I share these news with the people I have in mind so they may also log in there no only their certain joy with this new feature but also any other suggestions they may have in terms of getting Ricardo’s pronunciation right.

Should I take the action or you would like to do it?

[karl_lew]

@Gabriel_L

Arahant should sound like this

Oops. I forgot the second “a”. I can’t figure out how to shorten the ending “e”, however. AWS Polly is quite stubborn in this way.

• <phoneme alphabet="ipa" ph="'a.ɾa.han.te"/>

Here’s another attempt at Êxtase. I’ve doubled the “s” in an attempt to lengthen the initial vowel:

• <phoneme alphabet="ipa" ph="''ɛs.sta.se"/>

[sabbamitta]

@Gabriel_L

Yes, arahant will be a term used in the translations some times.

@Gabriel_L

we start a new topic just to capture any further pronunciation issues I or anyone else finds with Ricardo’s reading of the suttas

Good idea! Please go ahead.

[karl_lew]

@sabbamitta

Please go ahead.

Um. Tracking multiple threads of corrections would actually be harder for me.

For releases, we typically put together a Release Notes summary for succinct and broad distribution.

[sabbamitta]

@karl_lew

Tracking multiple threads of corrections would actually be harder for me.

Don’t worry. I can just check them and make issues for the several problems or post in our developer thread whatever you need to know. I understand that it’s just easier for the Portuguese group to have one particular place to go, with no ohter stuff posted in-between.

[Gabriel_L]

Done.

SC-Voice: feedback for fine-tuning Portuguese TTS voice (aka Ricardo)

As discussed in the topic https://discourse.suttacentral.net/t/wanted-translator-for-sc-voice-interface SC-Voice is being upgraded to allow the reading of suttas found in Portuguese language in SC. This topic is created to capture any feedback and suggestions speakers of Portuguese may have in terms of how to improve the pronunciation of by the text to speech (TTS) service, aka Ricardo. At this stage the Portuguese TTS service is found at the staging server. For reference, see below link to li…

@karl_lew

If it is not too much trouble lets continue our conversation on how to get êxtase and arahant pronounced right over there?

[sabbamitta]

@Gabriel_L

Happy to help! Could you kindly isolate and share the file to be translated as we did previously?

Here you go: language Wikis.zip (9.9 KB)

Where there are screen shots, you can if you like make the respective shots for Portuguese setting. Where there are examples, just choose appropriate examples for your language, if possible from the new segmented translations.

Thank you!

[scatterbrain]

@karl_lew

That is indeed how Voice pokes and prods the AI voices into something resembling what we want. If you are inclined to delve into correcting the pronunciations for the translators using Polly SSML, your help would be much appreciated. It takes a discerning ear and a good deal of patience with trial and error. The IPA is not always honored phoneme by phoneme. The AI tends to introduce contextual bias that will cause the same phoneme to be pronounced differently according to context. Would you be interested in helping with the SSML?

Yeah, I was 99% sure you were already using it, just mentioned it on the 1% chance you didn’t

I’m not very good at the ear part, sadly…

But, if you need someone to make a quick script to add SSML to certain words, in case you need something more complex than Find and Replace, I may be able to help with that …

[karl_lew]

Thanks to @Gabriel_L, who found a website that ALL translators can use to help out with Voice pronunciation corrections. This website uses exactly the same AWS Polly as Voice, so it is perfect for the job.

Feel free to use this website to experiment with alternate phonetic spelling for mispronounced words. We’ll simply incorporate your IPA corrections into the customWord vocabulary for your language.

For example, using this website, Gabriel and I figured out the Portuguese pronunciation for “arahant” after several iterations:

• <phoneme alphabet="ipa" ph="'a.ɾa.han.t͡ʃ"/>

This is an exacting task and it really works well to have many helping ears. It can also be a lot of fun to learn new ways to hear.

[Dhammanando]

Hello again,

@sabbamitta

In a term like this

foundSuttas: "Found A_RESULTCOUNT suttas",

(line 185 in the file) the bit in capital letters, “A_RESULTCOUNT” does not need to be translated. It tells Voice that in this place the number of search results should be inserted.

Does the rule of not translating words in capitals apply to the whole file or only to the aria section? For example, in the non-aria section, line 65 has:

contentUpdaterWarning: ‘WARNING: this task may take hours and may update nothing.’,

Should the word in bold be like this:

contentUpdaterWarning: ‘VARÚÐ: þetta verkefni gæti tekið nokkrar klukkustundir og gæti ekki uppfært neitt.’,

Or like this:

contentUpdaterWarning: ‘WARNING: þetta verkefni gæti tekið nokkrar klukkustundir og gæti ekki uppfært neitt.’,

??

[sabbamitta]

@Dhammanando

contentUpdaterWarning: ‘ WARNING

Yeah, of course, WARNING should be translated!