Kanji in ”type what you hear" sentences
Occasionally there are these "Type what you hear" (Speech-to-written) sentences. These are great fun and a good challenge - except for the fact that they almost never accept any kanji. What makes it even more frustrating is that the Report-option does not allow reporting "My answer should be accepted" for these type-what-you-hear questions. Why is this?
I always use my keyboard (using my laptop) rather than the word bank, because I find it much better to learn to type the Japanese words myself rather than rely on word bank. However, my computer will suggest and auto-replace kanji and I WANT IT THAT WAY because that way I learn kanji and I train myself to get used to the kanji.
However, for type-what-you-hear I need to manually ensure that no kanji slip into my sentences, because specifically no kanji are accepted for these. It's annoying, counter-productive and anti-learning.
So I have two requests:
Duolingo, please add kanji to type-what-you-hear sentences.
Duolingo, please allow us to report "My answer should be accepted" for type-what-you-hear sentences.
Thanks for listening.
What's worse is that sometimes they do require specific kanji, so you can't even get away with just typing it all in hiragana.
Same problem. I almost always see these listening exercises in timed practice too and have to skip them completely because I just don't have the time to go through and adjust everything that does or does not take kanji. It's frustrating there's no report option to fix it.
I suppose there not being a 'My answer should be accepted' option is actually the reason for the options with more Kanji not being in the answer bank.
I suspect that the issue is not that they're not in the answer bank. Rather, the issue is probably that the "Type what you hear"-type questions don't look through all the back-translations; it only looks at the source sentence. (I don't know this for certain, but it seems likely, given how the Incubator handles sentences.)
This is presumably because in most languages that Duolingo teaches (and especially the European languages that it started out with), the majority of the back-translations are completely different from the source sentence, rather than just being different ways of writing the same sentence.
For example, in French, you might have the source sentence:
Avez-vous une voiture?
With the first-listed English translation:
Do you have a car?
This sentence can then have many possible back-translations:
As-tu une voiture?
Est-ce que vous avez une voiture?
Est-ce que tu as une voiture?
Avez-vous une auto?
However, it wouldn't make much sense to provide any of those back-translations as an answer, because they aren't "what you hear"; sometimes they're interchangeable with the sentence you heard, sometimes they differ significantly, but in any event they are different sentences.
This is different from Japanese, where the sentence
may have the back-translations
which are all pronounced the same way, but written out differently, so ideally they should all be accepted. However, because Duolingo is modeled more on European languages, the system is not adapted to this.
I don't think there's really an easy fix for this; distinguishing between different readings vs. different back-translations would require significant changes to the way Duolingo handles sentences.
However, I think the best thing would be to allow all back-translations as valid answers for Japanese "Type what you hear"-type sentences. Sure, this would allow for some incorrect answers, such as replacing 映画 with フィルム, but I don't really see how that's a problem; it can't be "exploited" to make the problem easier, because rephrasing the sentence still requires you to have understood the sentence in the first place. In fact, it could be seen as an added challenge for ambitious students.
I don't know... French has (had? I haven't used it in a while) type-what-you-hear exercises that accepted homonyms like "elle tombe"/"elles tombent". So, it may be more complex in some way that's opaque to me, but there are already European languages that can accept multiple valid transcriptions for this type of exercise.
From a discussion started by a Norwegian contributor, apparently to add additional accepted responses in a write-what-you-hear exercise contributors have to send each sentence to staff as they're unable to do it directly in the incubator themselves.
I suspect the Japanese team is doing this, but whenever you've got to get staff to do anything, much less literally thousands of things, there are going to be delays. Perhaps they're implementing / hoping to implement an algorithmic solution. It would obviously make tons of sense.
I would speculate this issue might be part of the reason the course didn't seem to have any write-what-you-hear exercises until quite recently.
Ah, interesting; thanks for the info! The only course I've helped develop didn't have audio, so I'm not quite sure how this works.
Perhaps the audio questions have their own submenu, then, similarto image recognition and "Fill in the blank"-type questions. That might allow contributors to add alternative answers, but they may not carry over automatically from the full question (since "elle tombe" and "elles tombent" wouldn't actually be back-translations to the same sentence). In that case, pointing out desired alternative answers in the forum could still be useful.
File a bug officially through the website. And put what you wrote in the attached discussions. It's a valid frustration. It's getting better but still needs a lot of work and I think that requires user input, bug reports etc.
But I do feel your pain. Sometimes when I do Japanese lessons I'll get frustrated and walk away, then come back later when I have more time. Hang in there.