How to process natural speech?

Question

Machine learning

I am writing a bot for conferences (general text chat).
I want the bot to be able to perform some functions:
1) Do not break if the command was not pronounced letter by letter
For example:

Bot, tell me the weather

tell me the weather, bot

By the way, bot, what's with the weather?

2) Distinguish between the mention of him in a conversation, and a direct appeal.
For example:
Do not respond when:

the bot still can't upload pictures, i'll fix it soon

React when:

Bot, tell me the weather

In fact, the first point is much more important for me, and the second one can be dispensed with. I want the bot to be able to understand the call not only when the message starts with his name, people can write: "Hey bot", "botbotbot", "Boooot", etc.
It is also necessary to throw out everything superfluous from the speech and normalize it, to understand that "cat picture" and "cat image" and "cat photo" are one and the same command
. And the question is, with what help to do this? (because I definitely cannot write such an algorithm).
Is there something ready? Initially, my goal is in the functions of the bot, but I would really like to give him the ability to understand the natural speech of people, if this turns out to be too difficult, then it's not destiny.
I ask for articles with sources, links to repositories or libraries.
Thank you.

Answer 1

I ask for articles with sources, links to repositories or libraries.

Next! ;)

Answer 2

https://ai.google/

Answer 3

If we are talking about the Russian language, then read about stemming and lemmatization.