What does the architecture of a text chatbot look like?

Question

Bots

Hello

The task is to develop a text chat bot
It must understand requests from a certain area (selection of clothes) in natural language, requests like:
- I want a white dress;
- pick up an evening dress for the theater;
- Choose running pants;
- I need a light outfit in which to go to a club
, etc.

What might the architecture of such a chatbot look like?
So far, I have come to this option:
The bot interface receives a user message -> NLP -> trained neuron -sparql query-> RDF model (vector),
respectively, after the request is completed, we get one or more subjects (clothes models) that we send in response to the user

Can you please tell me how such a bot should / can be built and how correct is the above architecture? In what directions should you dig?

Answer 1

A network of synonyms with "weights" of "chains" of objects and their properties. (just connections, not a neuron)
Everything is extremely simple here, because no understanding of actions and relationships is required.

Answer 2

There is no single architecture. Each developer creates his own architecture and everyone has different technology stacks.
There is no single answer.
The architecture resulted by you has the right to existence. The rest is a matter of implementation.