If you work for a non-profit that serves speakers of languages other than English, you may find yourself using automatic translation to bridge the language barrier. It makes sense—machine translation is fast, free, and supposedly “as good as human.”
This post is not meant to dissuade you from ever using automatic translation. Instead, I would like to encourage you to use it appropriately and consider what alternatives may be a better fit in some scenarios.
How to Make It Work
Machine translation may be fine for casual, low-stakes communication that will not be publicized. Some examples may be friendly conversations and small talk, emails, directions, getting the gist of a website for your or your client’s reference. If you go that route, here are some things to consider.
Remember that Google Translator is not the only machine translation provider out there. While this service might have improved over the last decade, the output will vary depending on the language. For some language pairs, you may want to use a different provider, for example DeepL. Make a point of reading the provider’s confidentiality policy. Will it store your content on its servers as training data? If so, this option will not work for sensitive and confidential information. (Disclaimer: I do not represent or endorse any particular automatic translation provider.)
Ideally, have someone check if the resulting translation makes sense, especially if this is something your clients will refer to. Today’s automatic translation providers have shifted to neural machine translation, which may do better on fluency than prior iterations but worse on accuracy. This means that the translation may read smoothly but say something different from what you originally meant.
Write or speak in a straightforward manner that will be easy for machine translation to parse. Avoid metaphors, buzzwords, references to institutions and legal frameworks particular to your community, arcane language, and anything that may result in ambiguity and confusion. As of the time of writing this article, DeepL translated “The governor dropped the ball” into Russian as “The governor threw the ball” (Google translated the phrase literally, not picking up on the figurative meaning).
Automatic translation was not designed to be a language-learning tool. If you do use it to help your clients figure out how to say something in English, it is best to enter complete phrases rather than individual words. You will want to use this option with caution as the resulting translations could be misleading or inappropriate.
Consider Better Alternatives
Most importantly, remember that machine translation is not your only option! Here are some other ways of serving your clients.
Written Materials
Instead of using an automated translation plug-in for your entire website, which was not written with machine translation in mind, prepare a one-page summary of your organization’s services and contact information and have it translated by a qualified professional. That way, you can be sure that the information you are giving to your clients reflects what you had in mind when you put together the English text.
Check if the information you are trying to share with your client may already be available in the language of their preference. Many state and federal agencies and major institutions like hospitals will have language access plans, which require translating important documents into the top languages of their customers.
Conversations
Similarly, for face-to-face communication, many organizations and institutions are required to provide an interpreter free of charge. For example, if you are accompanying your client to a medical appointment, it may be worth asking the front desk employee when making an appointment what options are available. If no local interpreter is available for the client’s language, consider using video or phone interpreting.
Learning English
Put together and share with your clients a list of resources for learning English. This may include podcasts, apps, or dictionaries, for example Duolingo or Randall’s ESL Cyber Listening Lab. A monolingual English dictionary for learners, like the Longman Dictionary of Contemporary English, as well as a bilingual dictionary showing words in context, like Reverso, may also help your clients build their vocabulary and decipher new words. Finally, see what in-person resources may be available in your community. Public libraries or local places of worship often offer free or affordable ESL classes.
Machine translation is a powerful and ever-improving tool for bridging the language gap. However, this technology has its limitations and may not be appropriate for every task. Consider it alongside other solutions to decide which approach will best meet your client’s needs, protect confidentiality, ensure accuracy, and honor regulatory requirements.
Maria Guzenko 