Hello everyone,
and welcome to another episode of “A new feature.”
And yes, I know it doesn’t sound all that exciting after all the new features we already had this year. But this one is really really epic because it is about
practicing speaking with… AI
OMG, the buzz… it’s soooo strong :).
Here’s what’s going on… some of you might have heard of the company EF, Education First. Their main focus is
organizing language travels and programs around the world and they’re a pretty big company. And they have a small department that is building their own custom speech recognition AI system, that they’re offering to teachers so they can create cool interactive exercises.
Bob, the leader of the team (who used to be a fellow reader here), reached out to me a while back and told me about this sytsem and that I could include it here on my site.
Which was perfect timing, because I had actually been tinkering with that idea for a while and I had applied for access to Elon Musks Open AI in January and had I also looked into how I could integrate the Google speech recognition system into my site. But neither one is a feasible option, at this point, so when Bob told me about this, I was really happy because it was pretty much what I had been looking for.
And I’m really excited to show it to you today and I’m suuuuper curious for your feedback. And the team over at EF is, too, because they’re still very much developing the system and any feedback from actual students is super valuable.
Now, the system offers a few different modes. One is to just read out a sentence or repeat it another one is a dialogue where you can speak either side. But the one I found most interesting was to use your speaking to actually answer real questions, because that way, you have a real reason to speak other than just to get it graded, and you practice two things at once… speaking and some other part of German #efficient
So, today we’ll do a little exercise for the German past tense.
Here’s how it works:
I’ll give you a sentence in present and you’ll have to say it in past.
It’s multiple choice, so if you have no idea, you can just look at the options and read the one that you think is most likely.
Or you listen to my audio and then just try to repeat it.
Click on the blue mic to record and the same button to stop.
It’s only 10 questions, but it’s really more of a trial run. If you like this kind of exercise, I’ll definitely do more. I think for stuff like past tense, or creating questions or even adjective endings, this system is great and much better than clicking or typing.
But yeah, your feedback is what matters :)
Now, at the moment the system is completely run by EF, so you can’t save your recordings here and you can’t save the results in your profile, but you should be able to see these stats on the EF site.
Also, because everything is run by their servers, I don’t really have access to the layout or the “flow” of the exercise. If you have specific ideas there, all feedback is welcome.
Oh and one last thing… the AI is kind of in “nice” mode, so if it’s in doubt, it’ll “understand” in your favor. In fact, if you record and just say nothing, it’ll classify that as the correct answer (though you’ll get a poor rating for your pronunciation).
Anyway, enough talk from me, now it’s your turn :)
Viel Spaß!!
***
***
So… how’d you like it? How’d you do?
Do you want me to do more of these types of exercises? What would you do differently? And how do you feel about the AI? Is it too hard, or too easy? Leave all your thoughts and feedback in the comments and I’ll pass it on to the team.
I hope you enjoyed this, have a great week and I’ll see you next time :)
Hello from Berlin…I am just getting frustrated. I can’t get this exercise to even open and it sounds like fun and I could not make the recording work on the new set of exercises. I have a Mac and it seems that there is something that I need to be doing so it will work. I would gladly be participating and think it is a great idea. But alas…so far kein Glück :-(
Do you know how to do a hard reload of the page? That might fix it. Also, are you sure audio is enabled?
Es hat mir sehr gut gefallen, denn Sprechen ist was ich am liebsten tue. Elizabeth
Freut mich :)
I like your AI presentation. I hope you do more. 10 questions to answer are plenty. Like someone else commented, I also live in Texas and there are not many people to have conversations with here.
I’ll definitely do more. I’ll actually do a separate course soon, but that’ll cost a little extra :)
Too easy , es ist ganz einfach
Ha! What was your scores? Did you manage to get 100% ? If yes, then you’re definitely native level!
Ausgezeichnet!
Sehr nützlich, insbesondere die Möglichkeit, sich mit einem Muttersprachler (Sie!) zu vergleichen.
Danke :)
Sehr hilfreich ! Es wäre toll wenn Sie mehr Übungen machen können
Mach ich auf jeden Fall :)
Hi I’m Marcus, Ich möchte mich bei Ihnen herzlich bedanken specially to my sponsors so that i could access (for free) this valuable articles compact together jam-packed with bits of knowledge and information nowhere to be found in the internet. Again with sincere gratitude, Vielen Dank.
Fantastisch! It was great and fun. I love feedback. It’s hard to find someone to speak German too in Texas. So, an A. I. is cool and I got +95% after I figured out all of the buttons.
Haha, yeah, it’s not a native speaker but at least it gives us some sense of “being listened to”.
And it’s more accepting than the Google voice recognition for instance. That drives me up the walls sometimes, because it acts like it doesn’t understand the simplest stuff .
I loved it! please do more of them! My scores were kind of low but i think it’s maybe because i have learnt a lot of words in Salzburg pronunciation ^^
It also makes a difference how good the mic is and how far away you are from the computer/tablet.
I’d say more than half of this is about getting you to speak and getting the right answer… the pronunciation training is just a Trojan horse :)
I have tried this both in Firefox and Chrome, I allowed yourdailygerman to access the mic in both, but it still would complain:
:(
The whole thing loads the EF page in an iFrame, so giving yourdailygerman permission doesn’t extend to the actual app.
Normally, permission should be granted automatically, but looks like you have to do it manually.
Or you can take the exercise on the EF site directly:
https://speak.ef.com/go/theater/a-57vtde
let me know if that works.
Hello,
As usual, the “OK Boomer” is having trouble with the software. Or is it the hardware? Sigh. It always has to be one of us in the over 60’s, doesn’t it? I tried it with and without my mic/headset and all I can get is “Audio is too short or not loud enough. Please try again [Big, loud, “OK Boomer” sigh].” (Good AI, that even knows when to put in the sigh!)
Sorry! Lisa
Oops! I take it all back. I went away, came back, and the program is working perfectly. And I love it! I like the little recorder on the side of your lessons better, because I can hear my voice and compare it to yours, correct little wierd sounds that I am making (or not making) that don’t belong in (or are missing from) my (sorry-ass) attempts. With this AI program, I just get a score, but not a chance to hear myself. (Or am I behind the eight ball with the technology again? I can’t play myself back, right?) In any case, viel Spass!
Yup, hearing one’s recording is definitely something they should add (and will add, I believe). Glad you’re using my own little custom recorder. I have no idea how many of you use it because it collects ZERO data, so good to know it’s not going unused :). Have you noticed that you can slow down the playback speed also for your own recording?
Hmm, that sounds like the mic might not be active actually. What kind of device are you using?
Ich mache definitiv mehr :)
What a fun exercise! When it scored me low, I’d listen to your recording and try to mimic. The red/yellow/green coloring at the end was useful.
I think it would be cool to ask an open-ended question and allow us to give an unscripted response. In the transcription include the red/yellow/green for pronunciation.
Unfortunately, open ended is not possible with this AI. But you can try to practice with Google voice reconition if you want. Even Whatsapp has a transcribe feature that’ll write down all you say :)
But yeah, this one needs a sentence in writing to build a model of what it should sound like.
MIKE C
Please approve only one of my four posts,so that I can
communicate again. This one had me logged in, so I just
want you to be aware that I have an account already – but not
an imogee.
MIKE C
First time – so three entries
Ja, Ja,Ja mehr,mehr,mehr — ganz gut
Hahaha, nice. Yeah, I’ll definitely do more this summer. There are lots of things that are good to practice with this tool.
Solved issues by fixing my profile – vielen Dank
MIKE C
ist super! Vielen Dank
I stupidly didn’t realise there were choices not visible and had to scroll down the window, and also realised I was covering the microphone with my finger at points, but when I finally got sorted, it worked like a charm! Nice one!
This is cool, more of this