Computers are everywhere. We live in an age of cloud computing, smartphones, big data and self-driving cars. Surely we can take advantage of all this smart technology to automate the translation process? Unfortunately, it’s not that simple. Machine translation (MT) has made great leaps forward in recent years, but it still can’t replace human translators.
Human translators work by “decoding” the meaning of the source text, then “re-encoding” it in the target language. This requires a profound understanding of the grammar, syntax and especially the culture of the source language – and a similar grasp of the target language. This is something that computerized systems just do not – and arguably cannot – have.
The most common problems in machine translation relate to words with multiple meanings, non-standard speech, and translating named entities. Languages like English have many homographs (die = to cease living; die = singular form of “dice”), and machine translation can struggle with picking the right term from the context, as well as having trouble with idiomatic and ambiguous expressions. It can also stumble over word order and syntax.
Isn’t machine translation “good enough”?
Good translation is difficult. At the moment, machine translation can turn text into something where a native speaker of the target language might be able to figure out the approximate meaning, but it will almost certainly not be correct. Non-native speakers may well have a lot of trouble following the translation at all. The more different the two languages are, the worse the translation will be.
Machine translation might be “good enough” for some purposes, but any time your company’s reputation or image are on the line, you should seriously consider whether it’s what you really need. If the precise meaning of your text matters – for example, in medical documentation, legal texts or safety instructions – you really should be using a professional translation agency.
If you don’t want to insult your neighbors, anger your trading partners or end up on sites like engrish.com, you should seriously think about whether machine translation really is “good enough.” Otherwise you might end up trending on social media for all the wrong reasons. (The link is in Finnish, but try copying and pasting the headline and subheading into Google Translate to see how “good” machine translation really is.)