SSi internship opportunities? ![]()
That would be brilliant, but itās just me fantasizing at the moment ![]()
Do we have any feedback on the pronunciation accuracy from native speakers?
Also, do we know the dialect(s) the training material comes from? I do hear slight differences between the generated speakers, e.g., ceist a chur orm, but I could be making this up. ![]()
e: Iāve been using teanglann.ie (thank you @sara-peacock-1!) to compare words and how they sound across the 3 dialects and I believe the accent of the first speaker is Connacht. The 2nd Iām unsure about.
teanglann.ie is excellent, Iād say is one of the most best pronunciation sources on the internet for Irish, along with fuaimeanna na gaeilge
I can answer one of your questions. I remember reading on here that the translations were all checked by a native speaker.
I donāt think all errors are equally worrisome in that regard. There are errors that frequently fossilise and errors that donāt (and you couldnāt have reached as high a level in English as you have without knowing that as well as I do).
My impression from the Italian and Japanese courses is that the AI is good at avoiding those kinds of errors, perhaps because AIs learn languages in a very different way from people and tend to produce different kinds of errors.
Were there, really? ![]()
With the Welsh course, that I started with basically 0 prior knowledge of the language, I remember the effort of listening carefully to the sentences and wondering if that sound really had changed or even disappeared from that other previous sentence. ![]()
I donāt remember any comments about āmutationsā but I know my fried brain took a sort of break whenever the explanations and intros came so I may have forgotten them entirely.
However I know that thanks to this it was much easier for me to start speaking compared to many Welsh people who had maybe studied it in traditional classes for years and years, so I think thereās no real need to worry either way.
p.s. planning to come back and re-try the Irish (which I had attempted for the first time when I was in school and so. much into Ireland and everything Irish but never moved forward beyond a handful of words).
I can remember a comment along the lines of āyou might have noticed that the sound softened there - this is a special feature of Welsh, and we advise you not to worry too much about it for now, but notice where it happensā or somesuch. This was on the original āold courseā, so I donāt know about the new app.
I donāt know about the app either. I did the challenges, Southern version.
But I really donāt remember much of the comments (despite being without any doubt a very auditory learner, as I was never able to remember anything written unless i could āplayā it in my head or at least turn it into images).
I guess part of the reason could be I did Level 1 and 2 in just a bit more than a month - so there was probably no room for unessential words in English (my second language anyway) in my brain - although maybe I absorbed the idea of not worrying about whatever subconsciously. ![]()
I guess those who have learnt English as a second language are better equipped for this:
almost nothing we see written really matches expectations, when we hear it. ![]()
And even when we think weāve figured it out, we meet a first language speaker from a different area and itās all completely different again! ![]()
So with Welsh it just takes a little time to stop perceiving mutated versions as different words (with brain expecting different meaning for each one). But itās relatively easy, once you know that some letters just do that and that there is a pattern, which just becomes familiar after a while.
As far as producing the correct sounds, I admit that Irish and Gaelic seemed quite a bit more complicated to me in my mild and failed attempts at learning them.
But just like with violin, Iām sure, an effective method could bring big results pretty fast, also for adult learners.
Of course there is always people with a better ear and, letās say, ātalentā or maybe just more experience with languages, who are going to improve faster than others but I believe itās very much about finding the effective method.
We have a similar issue with Welsh, in the Census vs the Annual Population Survey. One asks āCan you speak Welsh?ā and the other āDo you speak Welsh?ā (canāt remember which way around off the top of my head) - a subtle difference that yields quite different responses.
I think most monoglot English speakers are blind to the difference between āI speakā and āI am speakingā that becomes very apparent when one starts to learn a Celtic language (I know I was before I started learning Welsh, despite being an English-language editor and knowing a couple of other, non-Celtic, languages).
As one who is about to go to the Gaeltacht for the first time⦠The āneedā vocabulary (cup of coffee, etc) is a great way in. When I rock up in the village, I want to show willing, and so starting a conversation about everyday things is an excellent way to show that I want to try.
Also, if my experience of the Bro Cymraeg is anything to go by, yes the everyday things can and do happen in English when someone doesnāt speak Welsh. But we really rather wish they didnāt have to.
From a language planning perspective, the ideal is that the language pervades every part of life from the mundane to the sublime, the bureaucratic to the poetic. It is just as important to be able to say ātwo pounds of rice pleaseā as it is to be able to connect to a national poet in the original language. Itās great to be able to appreciate the colour and nuance that makes every language unique, but without a community living their everyday lives through that language it will die.
Why havenāt those people living in the Gaeltacht learned the language? Is it because perhaps they donāt feel it is relevant to their everyday lives, and donāt have much interest beyond that? I heard from a Basque linguist yesterday that they have a saying (and I donāt know Euskera so I canāt tell you what it is) where they have to deliberately forget that they know the majority language in order to nudge people into speaking the minority language. Perhaps we need a bit more of that in our minority language communities?
And if someone only learns the mundane, the everyday - well at least theyāre contributing to the maintenance of the community (as opposed to its destruction), and perhaps thatās what suits their life, and thatās fine.
Good heavens, how long is the black belt section? I feel as if I will need to reschedule all my plans for the next 12 months! ![]()
I thought I was doing well, heading towards the end of Purple - planning to get to the end before my visit to Ireland at the end of April. But it seems that I might not make it! ![]()
(Only joking - that was a completely arbitrary target I set myself to keep motivation up - Iād happily go there now with what I have so farā¦)
As they say in martial arts, once you reach black belt youāve finished being a complete beginner and can start actually learningā¦
But yes, black is meant to take you far enough to be getting yourself in and out of conversations regularly, and thatās a big old lump of work. I do hope youāre starting to have some conversations, though!
I think itās a very good thing that there are increasing attempts at governmental and NGO level to get to grips with the implications of AI.
I also have a lot of faith in the human tendency towards diversity - I remember listening to conversationally fluent L2 speakers of English from China and Zimbabwe struggling to understand each other (despite both being understandable to me) and thinking that even as English was increasingly dominant it was already beginning to diversify.
I think community is key. Humans in community will always be linguistically creative, I believe, and thatās why itās such a common pattern for older generations to despair of the language use of younger generations. But if weāre isolated and trained on AI models, then I would expect to see monocultural tendencies - Iād also see the isolation as the dominant problem.
I was asked to accompany a young IT technician once on his first visit to a client. He was from Brunei and the client was Vietnamese. After the first 5 minutes I had no idea what they were saying - in English - but they understood each other perfectly! World Englishes in the making!
I remember talking to someone about āEnglishesā after my time in Zim and being told very sternly that āThereās no such thing as Englishes, youāre just making words upā ![]()
DunningāKruger at its finestā¦
Itās absolutely fine for you not to trust us - in fact, thatās probably a sensible starting point for any new social connections - I personally choose to trust in peopleās good intentions until I am given convincing evidence that Iām mistaken, but thatās just my personal preference. If you canāt trust us because we differ significantly from you in cognitive style, we will accept and respect that.
Weāve always believed in transparency, and will always answer questions as fully as possible (given time constraints).
Iām not sure we have current permission to share the name of our translator/proofreader - @Kai can maybe check?
In the current Irish course, we have a mixture of human translations from English, AI translations proofread by a human, and AI translations that havenāt been proofread yet (weāre working on an editing interface to allow people to suggest changes, but itās non-trivial). We primarily use Anthropicās Claude for translations and phrase generation, and either Azure or ElevenLabs for artificial voices. We are also working on an interface to allow people to contribute recordings - also non-trivial. I canāt go into any detail on what Claude was trained on because I donāt know. We chose it because it seemed in our testing to provide the best results of the current choices.
Please feel free to ask any further questions you have.