Deciphering language processing in the human brain through LLM representations