How to extract the answer to a question from the context like human answer

[dilolo]

Is there any way to extract a response from a given context and make the response human-like, like GPT does?
But without using LLM models