Technologies change all the time. It is sometimes quite laborious to keep up with the latest innovations especially when those innovations are so often unseen to the human eye. Sometimes the changes in the technology affect also the usage in a way that users can clearly see the new results, but mostly the improvements are not visible at all. Major and remarkable improvements in machine translation often fall to the latter group.
If simplified enough, today’s technological improvements are rarely more than changes in the code. Ordinary people won’t be able to see these kind of changes. Thus we need to find other ways to evaluate the improvements. As illogical as it is, we often only see and feel the design and make our inference based on that. If a tool looks simple its technology must be simple. If it feels precious it must be precious. If it sounds bad, it must be bad. And so on.
The invisibility is a real issue with the improvements in machine translation. With machine translation the cover has very little to do with the core. A simple package can hide an extremely advanced and innovative technology. The core and the technology are very important to the user of machine translation. It isn’t easy to build a machine translator from scratch. It easily takes a lot of work and a huge amount of data to build a machine translator. The development work that is made under the surface is the most important thing in machine translation. Still, the users can only judge the book by its cover, although the design is only a secondary aspect in the process.
It is a pity that the hard work dedicated to developing automatic translation goes so often unnoticed by the public. There are real innovations and ideas under the surface. Overall, translating is a very challenging field to technology. The translation quality will basically never be totally flawless for every text − no matter how many fixes are made or hours are spent. The user will only ever see the mistakes and imperfect quality. Translation quality is one of those things that are noticed only when there are errors. When everything works fine nobody pays attention to it.
Without the possibility to compare the before and after scenarios users may not even notice the improvement in machine translation quality. Because people can’t see what is done, the benefits of the improvements need to be sold to them. I’m not saying that the work for machine translation is useless. Quite the opposite! The changes and improvements just need to be communicated explicitly: the description on the cover should present the core perfectly.
Test qualified machine translation yourself. MT-Qualifier selects you the best available machine translation. Separately for each sentence. You get the best overall quality. Learn more here.