Online machine translation (OMT) is a fact of life for just about anyone who has any contact with a foreign language. What was once the realm of science fiction attracting not only the biggest names in technology, but it also now includes smaller startups that are beginning to make their mark in a fast-growing industry. Pushing this technology forward is its nearly universal accessibility: smartphones and cheap data plans mean that travelers are rarely without an internet connection, but even if they are, several of the major apps can now be used completely offline.

It is no longer realistic for language teachers to ignore online machine translation (OMT), or worse, to forbid students from using it. The following is a proposal for teacher-led student activities that help students build two key skills for using OMT productively: recognition of inaccuracies and the ability to correct them. Strictly speaking, comprehension only requires the recognition of what is specifically wrong with a translation, but exercises in fixing those errors will likely provide valuable engagement that would likely solidify understanding. 

Using OMT for production of an L2 text is a somewhat more challenging goal. Even if a learner knows that an L1 to L2 translation is not entirely accurate, it is likely that s/he may not know what part of the product is not good, so at that point, the product loses its usefulness. Recognition of what part is not good and the ability to fix it with an edit imply that the learner is at a level where s/he can recognize and correct a substantial range of errors, so we must assume that s/he would even be able to produce a good product without using OMT. If this is the case, then once again, OMT loses its usefulness: Using it requires editing, but not using it does not.

Taking a step back, OMT might possibly become a useful tool for production if an expert in the target language were guiding the actions of the learner. L2 to L1 ranking and editing activities could easily be used for L1 to L2 if and only if the instructor guiding the activity is an expert in L2. In particular, a full translation exercise using the products several different providers (and any other resources they can find) would allow learners them to compare products and use their intuition to pick and choose the best parts of each. Consulting other resources like online dictionaries, grammar guides and concordance or corpus tools (including Google n-gram viewer) would allow them to further enhance the final version.

Grammaticality judgments and fixing incorrect sentences are staples of language learning and textbooks have traditionally provided examples that follow the curriculum. Unfortunately, created sentences often lack a connection with the mistakes of actual learners, which give likely foster intuitions that textbook authors may not have (yet). In the same way, real examples of OMT errors may help to raise users' awareness of its strengths and weaknesses, again, in a way that teachers may not (yet) have the data that would be useful in intentionally designing them.

It is important to note, however, that it is crucial that these activities are a repeated, regular and meaningful part of the course curriculum. The goal is for the students to actually go beyond awareness to mastery of the ability to use OMT as a tool for language learning, much in the same way that a dictionary or grammar guide is used. It is strongly recommended that the instructor both guide the students to mastery of the target language and also openly participate in this exploration of the limitations and possibilities of OMT. This technology is constantly evolving behind the rather opaque barrier of patents, corporate strategy and, of course, the real limitations of even the experts' understanding of artificial intelligence and the nature of language. Therefore, instructors should constantly be alert for changes in the products of each provider and be willing collaborators with the students as they try to forge a path forward.