When comparing human vs. machine translation services, human translation services are always going to come out on top. While machine translation may be cheaper and faster, it also has significant accuracy issues that lead to consequential misunderstandings. This has resulted in machine translation developing a reputation for being low-quality, causing businesses to generally avoid it.
Most people who compare human vs. machine translation services typically see it as an either/or situation where you must choose one, but this is a misconception. Instead of weighing human vs. machine translation services against each other, you should look at how they can work together to improve localization strategies, accuracy, and productivity.
The Evolution and Infamy of Machine Translation
Machine translation (MT) dates back to the 9th century, when a cryptographer named Al-Kindi developed statistical analysis processes that make up the cornerstone of most machine translation programs today. Despite its early emergence, MT processes didn’t grow particularly sophisticated because they tended to center on word-for-word changes without the consideration of context.
MT was often found to make preposterous and embarrassing mistakes. In the worst cases, it would create content that was racist, sexist, or otherwise offensive. There’s a particularly good example of this occurring with Facebook’s platform.
In 2018, Indonesia suffered a major earthquake. Indonesian Facebook users marked their safety during the event with the word “selamat,” which roughly translates to “safe.” However, in Indonesian, it can also mean “celebration.” Facebook’s machine translation program inferred the latter meaning and marked many users’ posts regarding the earthquake with balloons, confetti, and other celebratory animations. In the face of a disaster that killed nearly a hundred people and devastated the region, the blunder was viewed as incredibly insensitive.
That’s one of the main problems with MT—when it goes wrong, it goes horribly wrong. While it has grown by leaps and bounds as more companies find ways to program computers to understand context, it’s still rife with errors. In most cases, human translators will still have to correct about 10% to 20% of machine-translated content to render it readable and accurate. However, when not treated as a solution but instead an assisting tool, MT helps make translators more productive and improves companies’ results.
Looking Beyond Human vs. Machine Translation Services
When faced with the question between human vs. machine translation services, the human translation will undoubtedly provide you with better results. However, instead of seeing it as an either/or scenario, you should delve deeper into how MT can improve the translation process and enhance the overall internationalization return on investment (ROI). It accomplishes these feats by:
- Enhancing translator productivity: MT and humans work together to improve the translation results. The MT completes the first draft, and then the human translator goes in and resolves any issues or errors. This tiered process improves productivity and the quality of translations. Humans can fill in the gaps that machines cannot, given their ability to infer context and use critical thinking skills to make the best linguistic decisions. Meanwhile, MT handles the computing power needed to complete the translation rapidly.
- Providing insight into demand: Every company has some low-traffic content that’s not vital to the website. An excellent way to gauge demand is to have this low-traffic content machine translated and reposted with a disclaimer. The disclaimer should disclose that the content is machine translated to account for any errors that may have occurred. Traffic to the newly translated content, user reports, and interactions can act as key indicators of which markets to target with more robust localization strategies.
- Testing applications: Translated content can wreak havoc on apps, causing uneven breaks in text, damaged code, or missing content. Testing how an app will react to translation is the most strategic way to avoid this, but you shouldn’t have to pay for translation when it’s just being used for testing purposes. Instead, MT offers a low-cost way to get translated content and then run it through your app. This method provides you with tangible evidence of how your app will perform, allowing you to make the necessary updates to ensure your new market rollout is seamless.
Bureau Works also leverages machine translation as a quality indicator, though this is not standard in the industry. Our localization management platform is transparent, so we can see how much time our linguists spend correcting translated content and making changes. This process helps us keep an eye on translators’ productivity and gain precise performance metrics.
We don’t view the process as human vs. machine translation services. Instead, we look at how the two can work together to create superior results. By integrating human intelligence with the speed of machines, it’s possible to improve the quality, accuracy, and turnaround time of translated content simultaneously.
Bureau Works doesn’t compare human vs. machine translation services, but instead integrates them, combining our linguists and our technology for the best possible results. Contact our team to learn more.
February 2, 2021