This could potentially be built into an e-reader's internal dictionary - selecting or hovering over a word or phrase could bring up (i) the internal MT's best estimate of the appropriate translation in context, and (ii) alternative meanings for the learner to compare against.

I think that one fundamental point in learning any language, particularly for those who have been brought up speaking only one mother tongue, is first of all to understand that 'source term A = target term B' very seldom applies, and such a mindset often leads to confusion and frustration.

A system that consistently indicates 'this probably means X, but the word can also mean Y or Z depending on the context' could be a very useful deployment of MT technology in language learning.