How does AI speech recognition work? How will I be posting?

Just before you start, I want to quickly explain how i will be posting & AI speech recognition system.

1. I will be trying to write how it really sounds.

For example, character ㄱ can be either  K or G.

To pronounce it more fluently, it is better to use G sound in which I will be using through out my post. 

Interesting to note, reason why K sounds exist is that even though Korean speakers would say 구(Goo), non-Korean speakers would hear Goo or Koo ! I do really see why K sound can be heard. 

 But like I mentioned previously, I want you to sound more like Korean native, so I will be only using the G sound using the example above.

You are NOT wrong hearing different consonants compared to what is written !



2. Using A.I speech recognition as a guide line

Google is providing in development A.I for people to use. Since it is in development, it is not perfect but I have found that it does quite a great job.

I can definitely make every recognition page better but for the sake of constantly posting, I will leave it as it is. It will show you the results in the decimal points. 0.9 would be 90%.

I would highly recommend you to use speech recognition in the most quietest place possible.

The recognition bar would fluctuate constantly as it picks up every single sound through the mic.

This is why I would ask you to use this as a guide line only. Just because the A.I doesn't pick up, it does not always mean you have pronounced differently. 

If you are using this as a guide line, you can interpret the result like this.

After speaking a word, give it about 1 second for the A.I to recognize. The result will print out in decimal points for short period amount of time.

Bar %

0 ~ 0.4 (0~ 40%) -> Definitely try again speaking few times as the noise around you may have affected the recognition. If not, return to the lesson to see which part you may sound different.

0.41~0.7 (41 ~ 70%) -> Also try again speaking to see if it improves. If not, you are very close to sound like a native ! Maybe you are pronouncing as K instead of G like #1 example !

0.71~1.0 (71~100%) -> At this point, you are most likely to sound near perfect or perfect ! As the A.I is in development, there will be some margin of error. Great job and move on to the next lesson !




If you have any concerns, comments, or questions, feel free to leave it down below.

Comments

Popular posts from this blog

Lesson 8: ㅇ (Ieung)

Lesson 3: ㄷ (Digeut)

Lesson 6: ㅂ (Bieup)