Named Entity Recognition for a Statistical Bot

I have a task where I need to recognize specific entities that do not exist in pre-trained ones.

I need entities about searching certain statistical info like: “give me the registered cars in California for 2020” → California is obvious a location entity, but registered cars would be an entity of type “datasets” and there could be for example filter-variables like the type of cars (variables in the dataset). But also California is also an information about the entity of state-level (as opposed to federal-level or city-level etc.).

So, I’m having my clear lists of entities (datasets, variables, geographic-levels etc.) but wonder how I could create a named entity recognition for that. Or is maybe named entity recognition the right approach at all? Of course it could also be possible to just work with the keywords. But then how to get also elements like “What is population with an age between 20 and 40” → The relational elements are also improtant to get “between 20 and 40” etc.

I would be very grateful to know what you think about this, and if you see some possible solutions for it.

Thanks in advance!