At a translation & localisation conference in Tokyo last year, Alibaba’s outsourcing lead of the International User Experience Division was asked by an audience member if machine translation was an option for her and her team. She firmly replied with a ‘of course not!’. It was quality, not cost, which was the most important factor…and an LSP that could help her sleep at night.
It’s perhaps no surprise that in 2017 the Association of Translation Companies reported a consistent rise in demand from businesses for machine translation. As Silicon Valley propels the MT movement, businesses want to know: is machine translation the answer they’ve been looking for?
As companies grow across international markets, managing a growing catalogue of content assets across languages becomes increasingly challenging. Communicating with a global consumer base increases the number of languages and overall volumes. For investment companies working on time-critical international deals, or law firms working on transactions across multiple jurisdictions, the process of translation can become problematic and costs can spiral.
Despite the technology giants hailing the wonders of language-related AI, machine translation is not a one-size-fits all solution for global businesses. It can form part of a translation strategy, but it won’t be right for all types of content. For gist understanding, machine translation could be a rapid and low-cost solution; internal documents for quick review could theoretically be served by MT because the output can be considered ‘fit for purpose’. Content going into the public domain should be of a professional standard (research or creative content), however raw MT output is still wide-of-the-mark in this area.
There are several ways to deal with large-scale content. Through MT, human translation, or a combination of both. It’s important to consider the priorities around quality, speed, cost and data security.
For machine translation specifically, consider the following options and pitfalls:
What EVS Translations recommends:
A closed-system developed by relevant training data for your subject area could provide sufficient output for gist understanding. Keep in mind, however, that for certain languages or subject fields, data is so sparse that machine translation is still not an option. Post-editing by a human translator could be an option to further increase the quality of the final output. What’s more, if your business consistently has large volumes, your translation provider can work on customising an engine to improve the output quality for your business over time.
What EVS Translations does not recommend:
The temptation for many clients is to use open-source free tools available on the web. But, as they say: “If it’s free, you’re the product”. Since these tools are collecting your data for further use, they aren’t appropriate for sensitive information. You also aren’t taking advantage of using your own translations as training data to customise an engine and improve future output.
If your business is experiencing large-scale demand for translation, speak with EVS Translations in-house translation technology team and find out whether machine translation could form a part of your translation strategy. Call our UK team at +44 (0)115 934 5010, our USA team at +1 404-523-5560, or use our contact form.
