Pali language: Itthirūpaṃ, bhikkhave, purisassa cittaṃ pariyādāya tiṭṭhatī”ti.
English language: The sight of a woman occupies a man’s mind.”
However, the speaker says: The sight of a woman NOT occupies a man’s mind.
It happens to the Rūpādi 1-10.
Please help check and update the content of audios.
Thank you very much for your feedback! And thanks for giving the exact reference AN1.1:2.2, this is very helpful.
Could you tell us with which voice this occurred? I’ve just tested with Amy, and she doesn’t actually say any extra word (“NOT”), but she quite strongly binds the two words “woman” and “occupies” together, like “womannooccupies” or so. Is this what you mean?
I believe that is what is meant. It is probably a quirk of the TTS, possibly an adjusted mapping of either ‘woman’ or ‘occupies’ will do the trick. Maybe try adding a mute sound which would still force a gap? Maybe ː will work?
Not sure, but this link may be of help: Convert IPA phonetic notation text to speech
Hi Dungdoant, thanks for your compliments about Voice.
We can fix mispronunciations to a certain degree and have already done a lot of such work, especially for Aditi, the Pali voice. We consider it quite important that people hear Pali as properly pronounced as possible so they don’t learn it incorrectly. But as each mispronunciation needs to be addressed individually this is a lot of work, and for translation voices we only do it if it leads to misunderstandings. If it sounds a bit strange but the sense can be understood we usually tolerate the mispronunciation.
In the case you mentioned this might be a bit of a border case, so I’d perhaps leave it up to Karl to decide (he’s the one who does the actual work in this area).
For the (hopefully) near future we are planning some work on pronunciation again, and here you can see the list of items we have at the moment. It’s looong!
Thanks for your efforts.
In fact, I often compare Pāli text suttas and English translation. So it is almost mentioned in Pāli.
And about translation voice, that is all right. That is so kind to your team working for Budda, Dhamma, Saṇgha. Once again, your team is doing a great job that all most people are looking for true happiness.
By adding a phonetic exception, we can sometimes correct a single word with customWord pronunciation ɒkʊpaɪz
Given the large number of words in use, fixing word pronunciation is fairly labor intensive, so much so that we have had to prioritize fixes:
Semantic fixes are highest priority because they lead to misunderstanding the Dhamma
Semantic fixes for Pali are given priority over other languages because Pali is the root language
Non-semantic fixes will be addressed as time and energy permit. This includes fixes for “non-standard” yet clearly understood phrasing.
As Anagarika mentions, we do plan on addressing pronunciation further this year.
Thank you for your patience listening to Voice and letting us know about mispronunciations. Together we train the AI voices!
We do indeed use IPA to correct AI pronunciation. It’s a bit of an art since IPA is not exact and quite open to interpretation. Different speakers will speak an IPA phoneme differently. Yet hope remains and AWS Polly offers SSML for us to tweak pronunciation. Amy’s IPA lexicon is limited but serviceable.