I'm not sure if this is what you're looking for, since its more at the level of data and psychology than neurology, but I'd strongly recommend reading A Human's Guide To Words. Most of your question - "How do humans understand what words mean?" - isn't specific to text, and applies just as much to spoken language, sign language, and any other form of symbolic communication that uses the brain's native language capabilities. 

At some level, language is always metaphor, where meanings are inferred by cache lookups searching only a few steps away in a very highly branched tree in concept-space. It has to be something like that. Poetry is just one or two more inferential steps removed from "literal" meanings, so we read it a little slower and still miss some of the richness of how all the concepts relate to each other.

I do wonder how much interpretation of written language specifically varies from person to person and language to language, but that's a separate question about how the brain relates written squiggles to the concepts it already has stored as "words." For me, I read aloud in my head, and thereby connect letters to spoken words. This... did not work for me when I was taking Mandarin classes and reading characters. Too many homophones. My teachers told me they never do this. For them, characters relate directly to words which have meaning, and spoken syllables relate to words which have meaning, but neither "are" the words. I don't know how common their experience is for anyone else speaking any language. I would be curious if deaf people reading English experience something similar, and if deaf people have an easier time learning to read character-based languages or less inflective languages. Or if illiterate people process homophones differently.