Irish sounds missing.
I don't know if this a bug or not. So, I posted it here. I'll gladly post it where it belongs. I started doing the Irish course. Many of the non-English words don't have accompanying sounds. For example, when you select a picture to match an English word, the Irish word doesn't sound. How am I supposed to know how to pronounce it? With a language like Irish, it's extremely important to be inundated with the sounds of things, as the pronunciations as completely different than the usual Romanic languages. Spanish, French, and Italian have sounds on every exercise. Thanks
I don't know if this a bug or not. [...] Many of the non-English words don't have accompanying sounds.
Not a bug.
As for many other languages (and more and more with new languages being added), Duolingo didn't find a decent third-party Text-to-Speech engine to use. Therefore they paid a third-party company to record the audio for 3000 sentences(*) in the course. Consequently:
- No audio on each word and, consequently, no "slow audio" on any sentence
- Sentences not among those 3000 have no audio at all
So, I posted it here. I'll gladly post it where it belongs.
It belongs here. ;)
(*) I think for the more recent courses it's now 4000.
Thanks for the response. The audio doesn't seem to make sense, though, even taking into account what you said. There are exercises that have audio, but then the same word in another exercise doesn't have audio. They, obviously, have the audio file that word, why not keep using it? Also, if they have real people doing the audio, why not have course contributors do the audio at the same time that they contribute to the course? Or, as someone else suggested, they could include a pronunciation guide to make languages like these easier to understand.
The audio doesn't seem to make sense, though, even taking into account what you said.
I think it does. There is no audio on words, only on entire sentences.
There are exercises that have audio
So, one audio file for the entire sentence, not word-by-word.
but then the same word in another exercise doesn't have audio.
Because this entire sentence has no audio recorded.
They, obviously, have the audio file that word
Nope, only for (some) sentences. And they didn't implement to automatically extract from sentence's audio the word of each word. Which wouldn't always work for all languages (if you have phenomenons like the French "liaisons" for example).
Also, if they have real people doing the audio, why not have course contributors do the audio at the same time that they contribute to the course?
They declined the offer of various contributors from various courses. One reason is that they want all the audios being made by the same voice.
This is an exercise/sentence (consisting of "one word + final dot"), so it's still the case that only the audio for 3000 sentences/exercises have been recorded: the audio are associated to sentences/exercise in the database not to words.
When you have audio associated to words, then you have slow audio.
So yes it is true that the word eilifint has no audio associated to it, only sentences/exercise have (among which the sentence/exercise "Eilifint.")
there is absolutely no difference [...] period/full stop.
In the pronunciation no difference, but in the database there is a big one: which is exactly why there is no slow audio and no audio associated to the word itself (only to some sentences containing this word).
The word is one entry (which has no audio in courses without TtS) in the database while a sentence/exercise is another entry (which has audio for "TtS languages", and has or not audio for "non-TtS languages", depending if the sentence is among the 3000 ones with audio) in the database.
The issue/point here for having audio on words (as opposite to sentence.exercise in which category enters the "sentences composed only of one word), isn't the language itself but how is made the database.
To my knowledge, the slow audio option is only available in courses with speech synthesis; if that’s the case, [...] exercise is.
Yep, exactly what I explained in my first comment. ;)
The languages without TtS not having any sound associated to the words (=to entries that are each "word" in the database), it can't have slow audio on sentences/exercises.
It is certainly an exercise, but without a verb, there is absolutely no difference between the pronunciation of a noun with a following period/full stop and the pronunciation of a noun without a following period/full stop.
To my knowledge, the slow audio option is only available in courses with speech synthesis; if that’s the case, then the Irish course would not have a slow audio option, no matter what the content of an exercise is.
So if a course with recordings had a “word” exercise in the database with an associated recording, e.g. if the Irish course had an eilifint “word” exercise with a recording instead of (or in addition to) the Eilifint. “sentence” exercise with a recording noted above, the “word” exercise would have the option of slow audio?
PS: Please reply to one of my other comments that has a Reply link, so that I’ll receive notification of your reply.
Thanks for the response. I've already seen the use of single word audio on exercises, but not on others. For example, sometimes when the word "Fear" comes up alone, it has audio. Other times, it doesn't. Also, NONE of the choose the picture exercises have audio attached to them, which, as I'm sure you know, are single word exercises. Also, I've heard, at least, two distinct voices while doing lessons, a man and a woman, and that is in multiple languages, not just Irish. I know what you're going to say, "that's 1 man and 1 woman". It doesn't matter. It proves that it isn't just one voice. Either way, who cares if the audio isn't always from the same person? Are you only going to talk to only one person in that particular language ever? It would be far better to hear pronunciations from multiple people so as to get a variety and a better understanding, because that's real life.
I've already seen the use of single word audio on exercises, but not on others.
Yes, because the Irish course contains exercises/sentences that consist just of one word (plus a final dot in general) and the team/staff had the third-party company recording audio for that exercises. But the recording is still associated (in the database) to the sentence/exercise, not to the word.
For example, sometimes when the word "Fear" comes up alone, it has audio.
Also, NONE of the choose the picture exercises have audio attached to them
Yes, because the audio aren't associated to the words in the database (but to sentences/exercises) and that Image exercises are, in the database, child-element of the word. So the image exercise hasn't the information about an audio to play in its "parent" element in the database which is the word.
Either way, who cares if the audio isn't always from the same person? Are you only going to talk to only one person in that particular language ever? It would be far better to hear pronunciations from multiple people so as to get a variety and a better understanding, because that's real life.
You'll have to ask staff. As I said, we were several contributors of several courses to offer to record audio but staff declined.
So, why not put it there?
- the database doesn't contain an audio file for each "word-entry" in the database: Duo sent a list of exercises(*) to be recorded and it has been recorded as one thing, not word by word. So the info "audio of the word" doesn't exist.
- because Duo didn't make the system like that: audio associated to "word-entries" come only from TtS.
(*) and, yes, there are few exercises that are composed of only one word, so the audio for the "word-entry" would be the same as the one for the "sentence-entry". But they are very few cases and Duo probably doesn't want to start having to manually change few things: if it were scalable (to all “word-entries" they probably would consider it).
Other languages do.
Other language with TtS do have audio associated to "word-entry". It doesn't come from any recorded audio of a "sentence-entry". So not comparable, you can't compare how things work for languages with TtS and languages with only "3000 audio files", each one being on one entire "sentence-entry" of the database.
it seems a simple task to me. If there is an audio file for an exercise with just a single word in it, that same file should be used in ALL of the other exercises with that same, single word in it.
But the other exercises contain other words... so you can't play this "one-word" audio for them.
Note: Audios on exercises aren't the sum of several audio files (one for each word) played. It's one audio file for the entire exercise that is played.
Exactly to avoid strange things that would happen (or even wrong pronunciation in some languages in which you have liaisons for example) if you played several file (one per word) in a row.
Aside from all of that, by posting here, I was under the impression that I WAS asking "the staff." That's why I asked if this was the right place for this.
There is no place to "ask staff". Some staff members sometimes come along here but very rarely (well, at least they very rarely post/comment).
Troubleshooting forum is the place where users help/answer each others questions.
Am I wasting my time?
If you don't want an answer from users knowing a how works the system (from inside, from the incubator) and only want an answer from staff, then yes you're probably wasting your time (and us too ;) ).
Were there two distinct voices in courses that use recordings rather than speech synthesis?
[Message edited 2016-07-26 08:50 GMT]
I'm not sure any "non TtS" course have two voices.
And even for TtS courses, afaik, it's still not a common thing:
- French has it for sure
- Portuguese had it at least as an A/B test.
Not sure if the A/B test ended and if it did if it has then been released to all courses or to none.
Halfway through a lesson in French, I don't know how to link to it, but the exercise for "a plus tard" is a male's voice. The others on the lesson, so far, have been female. Not that I have a problem with two voices. I think it should be more.
Irish has never had a male voice. I’d thought that French was one of the courses with speech synthesis rather than recordings.
I’d thought that French was one of the courses with speech synthesis rather than recordings.
Yes, it is, sorry, I didn't see the "with recordings" in your previous message. I'll edit my other message.
French is a course with speech synthesis, not recordings; the speech synthesis that Duolingo uses offers a choice of voices.
Thanks, but I was thinking more specifically and more English alphabet. Like when you hover over the word or something. I saw another place where someone suggested "buachailli" be shown like boo-kal-ee. That's all. Learning Irish is different than learning things like Russian or Greek. Those have different alphabets, where Irish has what looks like the English alphabet, but with different phonetics.
The problem with that approach is that there are many phonemes in Irish that aren’t in English, so e.g. transcribing buachaillí as “boo-kal-ee” would miss that a broad Irish CH ought to be pronounced as it is in German Bach rather than like English K.
Yes, use of “kh” would avoid potential confusion with the “ch” of “chair”. How does your Russian language book transcribe other sounds that aren’t in English, such as the vowel ы, the consonants ж, ч, ш, and щ, and the combination хг (e.g. in бухгалтер [“accountant”])?
If i remember right, ж is zh, ч is ch, ш is st, and щ is sh. I don't have the book with me, but I think that's what they did. I think ы is something like voy.
I found Lesson 1 + Appendix I of Ó Siadhail's Learning Irish book have been super helpful to study Irish pronunciation. He gives a lot of examples for each combination of letters, and uses phonetic alphabet as well. The sound files that go with the book have been helpful by breaking down most syllables into standalone sounds (i.e. one syllable words).
English “zh” (IPA /ʒ/), “ch” (IPA /tʃ/), and “sh” (IPA /ʃ/) are only approximations of Russian ж (IPA /ʐ/), ч (IPA /tɕ/), ш (IPA /ʂ/), and щ (IPA /ɕː/); these Russian phonemes don’t exist in English (and these English phonemes don’t exist in Russian). Similarly, ы (IPA /ɨ/) isn’t found in English; I’m not sure where the “v” in “voy” comes into it.
The same would occur with a number of Irish phonemes, for which English transcriptions would only represent approximations of the sounds.
I would say the "English" symbols, in conjunction with the Russian/Irish sounds, makes it much easier to speak and understand.
It’s certainly easier for a native English speaker to pronounce those substitute English phonemes than it would be to pronounce the proper Russian or Irish phonemes. As far as an aid to understanding, I could see that in the short term, but if one intends to pursue study in either of these languages, one might as well become familiar with the phonemes and the orthography of each of these languages sooner rather than later.
Before the new audio was introduced there were other voices (spoken by two(?) elderly ladies, as far as I can tell). Was anything wrong with that audio?
If staff decided to change the voice(s), it's indeed likely because there were issues with it(them), in which case they relied on the contributors to tell/confirm the voices were bad (or at least that it could be better).
The previous set of recordings were done by non-native speakers of Irish, and their pronunciations often did not accurately reflect those of native speakers. The Irish discussion forum contains many comments noting where the older recordings had incorrect pronunciations.
On a side note, how do you get to level 22 of a language? I completed Italian and I'm only got to 12.
Where is the account immersion?
Sorry, not following you.
The second condition to access immersion I mentioned it that you account ("Altair0315") has the feature enabled by Duo (which may not be the case: there is an A/B test running on that feature).
But if you have already seen the tab "Immersion" in the blue bar on top of Duolingo pages, then it means you account does have this feature.