Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Show HN: I built an AI language teacher to get you speaking

gliglish.com

Post reply on HN
Print view View original

71–80 of 112 posts

Re: Show HN: I built an AI language teacher to get you speaking

#71

Post by ihm »

fabiensnauwaert wrote: post #59
ihm wrote: post #49

Mexican Spanish didn't work at all. It asked me "Como te llamas" and I replied, "Me llamo Isaac", which got speech-recognized into some French sentence. I tried again saying "Si, me llamo Isaac" and it got recognized as some gibberish like "C. Mi anno E Sack" or something.

Did you try disabling multilingual speech recognition? Speech recognition is multilingual by default, which means you can ask a question in any language. The tradeoff is that your pronunciation may be misunderstood at times. e.g. English spoken in a thick French accent may be misinterpreted as French. If you turn off multilingual speech recognition you won't have such issue. Maybe I should have made it the default? A…

For what it's worth, I tried Spain Spanish afterwards and had no issues at all. My accent is pretty good (not native but grew up hearing and speaking Spanish as a kid, can sometimes pass for native if we don't talk enough that they hear my poor vocabulary).

Re: Show HN: I built an AI language teacher to get you speaking

#72

Post by fabiensnauwaert »

willsmith72 wrote: post #48
ozzeruk wrote: post #23

Okay, here's some brutal feedback, please take it with your best interests at heart. I am an English native speaker who has lived in Spain for 10 years, and has become fluent. 1) These speech-to-text models are poor when it comes to non-natives. This is unfortunate as the idea you had and the product you've designed could be incredible for language learning. However - it's a bit crap - sorry - I can speak Spanish wel…

Eh I disagree. It's not perfect, but that's just about expectation management. Users don't expect voice-to-text to be perfect - in fact, the past 10 years shitty experiences have been the norm. I think it depends on the level of mistakes the transcribing makes, but that's only going to improve with time with advances in AI and as the product evolves (using the context of the conversation like you say would be a great…

And, well… "glass half-empty or glass half-full?"

Speech recognition is far from perfect but even then it's incredibly useful. It CAN be infuriating (or downright hilarious). Hell, it's infuriating when occasionally a (human!) waiter switches to English after I mangled a sound in Hungarian, even though I'm C1/advanced in the language.

Problem or opportunity? Like someone else pointed out, the limitations of speech-to-text can be turned into an opportunity to improve one's pronunciation. It's getting extremely good for native speakers. As foreign speakers it's a chance to improve.

In any case, I'm sure I can add a layer or two to the code to reduce misunderstandings. This is actually exciting!

p.s. as mentioned in the OP, feedback on pronunciation is planned (actually in the works).

Re: Show HN: I built an AI language teacher to get you speaking

#73

Post by fabiensnauwaert »

lisasays wrote: post #26
ozzeruk wrote: post #23

Okay, here's some brutal feedback, please take it with your best interests at heart. I am an English native speaker who has lived in Spain for 10 years, and has become fluent. 1) These speech-to-text models are poor when it comes to non-natives. This is unfortunate as the idea you had and the product you've designed could be incredible for language learning. However - it's a bit crap - sorry - I can speak Spanish wel…

Exactly the level of quality we expect when the product idea starts out with "AI" and "scalable". And completely forgets (or doesn't bother think about) what beginning students actually need.

I learned for foreign languages and was obviously a beginner in each.

Not sure what hating on 'AI' and 'scalable' (your word by the way, not mine) accomplishes.

Re: Show HN: I built an AI language teacher to get you speaking

#74

Post by fabiensnauwaert »

IanCal wrote: post #40
knutzui wrote: post #28

Earlier quoted context omitted.

Slightly off topic, but I could imagine that what you are alluding to regarding the expectation of certain words or phrases depending on the context of the conversation could be used to improve speech-to-text models. The speech could be parsed into multiple options which can ranked by a language model with the conversation context.

Whisper takes a prompt as well, it would be a good idea to try that out.

It does and that's indeed Whisper I'm currently using. I do have mixed feelings about it:

- On the one hand, it performs well in so many cases… and having multilingual support built-in is great! - On the other hand: there's actually NO OPTION to Whisper to recognize just two languages (you either recognize ONE language or ANY language with it, which can cause issues depending on one's pronunciation and the language at hand.)

Will definitely turn OFF multilingual speech recognition by default, because the huge majority of negative reactions in this thread stem from this.

Re: Show HN: I built an AI language teacher to get you speaking

#75

Post by fabiensnauwaert »

grugagag wrote: post #42
ozzeruk wrote: post #23

Okay, here's some brutal feedback, please take it with your best interests at heart. I am an English native speaker who has lived in Spain for 10 years, and has become fluent. 1) These speech-to-text models are poor when it comes to non-natives. This is unfortunate as the idea you had and the product you've designed could be incredible for language learning. However - it's a bit crap - sorry - I can speak Spanish wel…

It completely misunderstood me as well in spanish, it actually inputted english instead of spanish. I like the idea but this is not working at all for me, at least for now. For the creator: Do not get discouraged, I hope you do get this working properly and see a lot of traction.

Did you try with multilingual speech recognition turned off?

Re: Show HN: I built an AI language teacher to get you speaking

#76

Post by zapperdulchen »

fabiensnauwaert wrote: post #66
jjkeddo199 wrote: post #20

Earlier quoted context omitted.

Getting English translation is a good idea, but it should be "hold down 5 seconds" to reveal -- Allowing it to be shown instantly might actually slow language learning. Users should be allowed to struggle a few seconds first (struggle === learning) before translating. Alternatively, it might be nice to be able to translate only individual words. If I understood 60% , inferred 35%, and was clueless about 5% of a forei…

Hey, thanks! :) Will add Japanese and Chinese momentarily. Which variation of Arabic would you want to learn? You make a good point. I might need to hire a salesperson or do a business deal for this.

I would rather invest in a Marketing person. A lot of stuff in the language learning space is sold thru content marketing. There are a lot of influencers. And they are very open to test drive stuff like this.

Re: Show HN: I built an AI language teacher to get you speaking

#77

Post by fabiensnauwaert »

emursebrian wrote: post #50

I am also working on an app for learning langues and I must say, you've built quite a lot in 6-7 months! Gaining and retaining users is always a problem that startups face. With language learning apps, it's hard because the marketplace is already crowded and learners will stop when they've learned a sufficient amount or just give up. Prioritizing marketing and promotion can be difficult because it's very easy to get…

Thank you for the encouragements!

> you're probably in need of content that users will feel compelled to share on social media and other places

That's a very good point.

> It was slow to send the recording then provide a response, the speed would make it impractical to use in a practice setting

Thank you for pointing that out. I haven't had any issues so far, and now realize the traffic PEAKED after this post. I should have scaled the server, my bad.

> I'd rather it ask me what I wanted to talk

Good point.

Re: Show HN: I built an AI language teacher to get you speaking

#78

Post by johtso »

fabiensnauwaert wrote: post #68
johtso wrote: post #25

What are you using for the speech-to-text? One thing that I really liked about Whisper was that I could mix english and Spanish and it would just work.. like "what does "una palabra española" mean?" and it would just work. Doesn't seem to work with this though.

It's Whisper as well and should behave the exact same way. Are you in the same setting (mic, ambient noise)?

what model? is it running one of the smaller models client side?

Re: Show HN: I built an AI language teacher to get you speaking

#79

Post by fabiensnauwaert »

plorntus wrote: post #37
ozzeruk wrote: post #23

Okay, here's some brutal feedback, please take it with your best interests at heart. I am an English native speaker who has lived in Spain for 10 years, and has become fluent. 1) These speech-to-text models are poor when it comes to non-natives. This is unfortunate as the idea you had and the product you've designed could be incredible for language learning. However - it's a bit crap - sorry - I can speak Spanish wel…

I also had difficulty getting it to understand me. Theres a couple solutions I can think of that may make this more usable: 1) Speech to text into an input field, allow the user to modify 2) I presume this is uses an LLM to generate the responses, submit the new text and give it the entire convo as context but initially ask it to "correct" the text to what would make sense in context based on similar sounding words.…

I should have disabled 'Multilingual speech recognition' by default. Lesson learned.

Re: Show HN: I built an AI language teacher to get you speaking

#80

Post by fabiensnauwaert »

GolDDranks wrote: post #44

This is a very promising platform. However, it misses some very crucial things about learning languages. Here goes: Learning languages is mostly about getting tons of input. I tried using this for a few minutes, but all the input I got was some lines of text. Also: People say that learners don't get as much chances to output as they would like. That isn't mostly true. And even farther from the truth is to say that ou…

Hey, thanks for the input and happy to have a healthy debate!

I'm a bit confused though:

> all the input I got was some lines of text

Didn't you hear the audio? Or what else were you expecting? I've experimented with the length of answers a bit and people are put off by long answers.

·

I think people overcomplicate language acquisition. Or at least here's my two cents from learning four foreign languages to fluency and now seeing my kids grow up trilingual. There's a natural language to learning a language:

1. Listen 2. Speak 3. Read 4. Write

Foreign language education has been switching the first two with the last two and keeps wondering why the majority of learners have such thick accents. (Meanwhile, kids who grew up with YouTube sounds great.)

You need input to observe and 'pick up' the language. But you need experience to grow confident and fluent. No speaking practice means no confidence speaking. e.g. if you never spoke on the phone after years of learning you're likely to get cold sweats on your first phone call for work.

Time and motivation play a big role. You just cannot squeeze the equivalent of years of learning as a kid into a couple of months (usually before giving up) as an adult. Making language rewarding (which means seeing progress and having fun) is critical.

·

Now my observation is that people just don't talk much in most classes and that people should be able to learn foreign languages in the most natural way possible: conversations (listening and speaking.) It's somewhat ironic that ARTIFICIAL intelligence is now enabling this.

I'll put my money where my mouth is: I'm still working on the product and when it's ready will use it to learn a language FROM SCRATCH :)

Post reply on HN