Isn't a plain text renderer as simple as "concatenate all the sentence structures in order" (pseudocode, I know)? Something like a tree traversal + StringBuilder or the non-.NET equivalent of it? So couldn't you write one based on the sentence object definition you are obviously already using? And wouldn't the large complexity difference be a reason that NLP oriented frameworks only seem to do parsing? (Please feel free to tell me off for not getting the question if you think this is approriate...) - Jarno Peschier