HOW GOOD ARE MACHINE TRANSLATIONS (PART 1)?
Many people think machine translations (MT) are the result of a simple process that involves nothing more than introducing the original text and submitting it to one of the available MT programs that are available on the market, and then using the results of this automatic processing.
The process is, in fact, a little more complex and time-consuming. Indeed, the many poor quality translations populating websites (especially e-commerce pages) are a good illustration of how, if you want a minimally acceptable translation, it is not good enough to put your text through “Google Translate”.
At translation agencies, when we speak of MT, we normally speak of “post-edition”. The tools offered by MT are just an aid that helps the translator save time. As an aid, they automatically produce a “raw” text that the translator will need to check, adapt and perfect in order to produce a final text that corresponds to the original in terms both of content (the appropriate terminology, reliability, etc.) and form (for example, adaptation to the conventions of the target language).
As with any other product, there are many MT tools, some better than others in terms of the technology or design upon which they are based (rule-, statistical- or neural-based technologies). The final quality of a translation made using MT will depend, first, on the quality of the automatic processing: the better the “raw” text, the less work there will be for the translator and the quicker it will be to edit. And if there is more time left over, there is more time to focus on working to improve the quality of the final text.
So, to sum up, machine translations will never do away with the need for a human translator to intervene. This is because, no matter how developed and perfected the machines may become, the text will always need a human hand and eye to ensure the quality each customer expects when purchasing a product.