So basically language decoding is multimodal like one of the other comments here said? Like in digital communication, perfect decoding needs appropriate SNR (context) in addition to optimal spectral efficiency (conciseness/preciseness of the language). Interesting research.