Formal evaluations of machine translation (MT) systems give a user a general sense of translation accuracies. For users with a passing knowledge of a target language, it is hard to establish reliability of translation results from their native source language. Have you seen any MT systems addressing the issue of confidence? Perhaps some interactive interface features? One of my students asked me: "How do I know when to trust MT? How do I evaluate the accuracy of the results?" Neither of the extreme answers seem satisfactory. Nor are they practical. You could say: "Well, you can't trust MT". Or you could suggest: "Consult a native speaker, or learn the target language". Is there an in-between answer that's feasible? Or, is the MT field simply "not there" yet in terms of making Watson-like goodness-of-fit judgements?

More Victoria Rubin's questions See All
Similar questions and discussions