AncaWe can trace the idea of Machine Translation back to the 17th century, in the work of René Descartes. But it’s the 1970s which saw Machine Translation used for its actual purpose, initially in institutions like the European Commission, and later at big corporations. The advent of the Internet sped up the evolution of MT significantly and resulted in advanced technologies like today’s Statistical Machine Translation.
In software localization, we can use Machine Translation (or Automatic Translation) in a number of processes.
Machine Translation as a pseudolocalization technique
Pseudolocalization means simulating the localization process before beginning the real translation work. It is done for testing purposes, without involving any human translators. Its purpose is to see what the localized software would look like in the target language. Pseudolocalization helps avoid issues related to text expansion, character encoding, string hard-coding and other aspects. Out of the available pseudolocalization techniques, Machine Translation is probably the best to imitate what would happen during the actual human translation process.
Machine Translation as part of the post-editing process
Although controversial among some translators, companies around the world are increasingly seeking the use of translation technologies like Google Translate or Bing Translator to localize apps or websites. They couple the input from these technologies with post-editing, with the aim to speed up translation turnaround time and decrease translation cost. There are different levels of post-editing, from light to full. Light post-editing only ensures readability and factual correctness of the translated content, while full post-editing produces translated content that complies with established grammar and style rules and terminology.
POEditor’s Automatic Translation feature
At POEditor, we offer our users the possibility to automatically translate software strings. With the Machine Translation APIs from Google (Google Translate) or Azure (Azure AI Translator), you can have the strings translated from a chosen source to any of the target languages available with Google/Bing.
Both Google Translate and Bing Translator use Statistical Machine Translation. This is a translation algorithm based on language pattern matching. While Machine Translation can provide satisfactory results when translating words or single sentences, it’s well known that translation accuracy can dramatically decrease with the increase in complexity of the text to be translated.
We always recommend using human translation when you localize a website or an app. This way, you’re a lot more likely to provide a high quality experience to the user. In case you’re on a budget and can’t hire professional translators, you can always use POEditor to crowdsource translation or maybe use a mix of Automatic Translation and translation crowdsourcing.