Yes, I do semantic chunk to better split the text into coherent chunks. I even use Llamaparse api (by Llamaindex) to parsing any type of documents, images or imported study material, with a very high quality results. Llamaparse is great!