Machine Translation (MT) has been a widely debated topic in the translation industry for years. Some argue that machine translation can never match the quality of human translation, while others know that it can provide a quick and cost-effective solution for certain business challenges. A lot of the analysis to determine if machine translation can be used instead of human translation revolves around whether the source content is "suitable" for MT or not. However, what if we've been asking the wrong question this entire time? What if I told you that all content is suitable for MT, regardless of what consultants might claim?
Machine translation suitability: predicting outcomes on a binary plane
To predict how well a machine will translate specific content, translation companies typically conduct machine translation suitability tests. These are designed to evaluate the quality of the MT output based on theoretical (or actual) input texts. For example, a translation company may conduct a machine translation suitability test for technical manuals to ensure that the MT output accurately conveys the meaning and intent of the original document, without losing any of the industry terminology, accuracy, and nuances without human intervention. If the MT output is not of sufficient quality, the translation company will choose to use a human translator instead or advise against using machine translation. Ultimately, the goal of these suitability tests is to determine the most effective and efficient way to translate a particular type of content while ensuring the highest possible quality. With this, standard matrices are created and it’s commonly thought that only specific content types are well-suited for MT workflows.
Traditionally, translation success criteria have been centered around three factors: quality, speed, and cost. Machine translation suitability tests are largely focused on quality, with the assumption that human translation output is superior. While it's true that human translation is currently unmatched in terms of its ability to convey nuances and creative expression, it comes at a high cost and is often too slow to keep up with the pace of modern content creation.
With the proliferation of digital content generation, the need for fast and cost-effective translation solutions has become more pressing. It's time to reexamine our assumptions and consider whether quality is still the most important factor in determining the suitability of MT. After all, if the ultimate goal is to produce an accurate and useful translation for business purposes, then speed and cost are equally important considerations.
Rethinking success criteria in translation
For some use cases like digital forensics, surveillance, complaint management or eDiscovery, quality is not the most important success criterion: time to information is. In these scenarios, human translation cannot operate at the scale or speed required - and MT becomes the only solution to the business challenge.
It looks like that we have been approaching the MT debate with a binary mindset. We've been assessing the output translation from a quality perspective and, often, the comparison to human reference translation has been the only factor that guided suitability tests. But what if we tried to frame the discussion differently? What if we asked, instead, "what do you need to do with the output translation?" and focused on the utility of the final business outcome?
When we engage outside of localization workflows, often the goal of translation is not to produce a perfect rendering of the source text, but rather to convey a message accurately and efficiently. MT can be a valuable tool in achieving this goal, especially when dealing with large volumes of content and tight deadlines. When speed and cost-efficiency are the main considerations, MT can be a game changer, regardless of content type.
The utility-driven approach to machine translation
What if we stopped thinking about content as being suitable or unsuitable for MT and started thinking about what we want to achieve with our translations? What if we asked, "what’s the goal of this translation? Who is going to consume it and what are their goals and expectations?"
By shifting our focus away from the content itself and towards our desired outcome, we open up new possibilities for MT. For instance, while I may not use MT alone on my customer-facing marketing materials because I expect that the content will resonate with my German market, I would use MT to make my support forums more easily accessible to my global audience. In this frame, content is irrelevant as machine translation can translate anything so long as it can read the source input text. The ultimate utility of translation is not achieving a linguistic outcome, it is achieving a specific business outcome.
The Bottom Line
Let's not get caught up in the false dichotomy of "suitable vs. unsuitable." Instead, let's focus on our desired outcome and use MT as a tool to achieve it. So, the next time you're faced with a translation challenge, don't ask yourself if it's “suitable” for MT. Instead, ask yourself what you want to achieve, and let MT be one of the tools in your toolbox. We will be looking deeper into the evolving perspective on MT suitability through a series of blogs to come. And, before we go, one more idea to think about: is there such a thing as overtranslation?
Are you ready to enhance your translation processes and gain a competitive edge? Discover how Language Weaver can transform your business and propel it towards success in the global marketplace. Read the 5 ways to improve your business translations with Language Weaver and unlock the limitless potential of machine translation for your organization.