How does this program transform text to vector which can be put to a lstm?

[DuckDuckBug]

I am not familiar with tensorflow, keras seems more friendly to me. I want to implement it using keras.
I have some question about it:

1. How does this program transform text to vector which can be put to a lstm?
2. Did you use embedding, how to get it?
3. How to calculate the rebuilt error?
4. How do you get the threhold?

As far as I know, http requests are not equal to nature language, so, I am very confused how you transform a one-hot char to a word-embedding vector?

[amurzina]

Hi!
We have quite detailed answers in that post AI Village.