How do pronunciation apps work?

Question

CMS

Hi friends.
So I became interested in how applications like Duolingo and others work.
What do they use for tests like:
Say this word in English, do you speak, and the application checks through the microphone what you said?
That is, in principle, there is software like Amazon that uses it.
I would like to know more about what they use? Which way to look?

Answer 1

This applies to sections of mathematics: digital signal processing (DSP, English DSP) --> pattern recognition (English pattern recognition) --> speech recognition (English speech recognition).
I think that it works like this:
1. capturing sound from a microphone
2. primary sound processing (cutting off the ends, removing noise, etc.)
3. extracting phoneme characteristics
4. comparing with the standard and getting the difference
Another link on the topic:
https:/ /habr.com/company/yandex/blog/198556/

Answer 2

Based on the fact that today even the leading and richest companies cannot make voice control 100% of high quality, then "pronunciation training" is fantastic.
marketing bullshit.
somehow approximately it works, but I would not learn from this, I would not set myself pronunciation without a living native speaker.

Answer 3

I am not an expert in this matter, but in my thoughts I would try to dig in the direction of digital prints. Like how shazam works.