While machine translation is a technical marvel, it appears that it is not going to push aside professional translation services anytime soon.
Let us find out why.
Even Perfectly Bilingual Humans Struggle to Put Out Perfect Translation
Even though professional translators are not walking form of dictionaries, they are better than plain old bilingual people in that they understand that any translation job demands more than just understanding of two languages. They try to do better than creating one-to-one equivalencies from one language to another. In crafting meaningful phrases and appropriate sentences, professional translators recreate language in another form so they have the same and equivalent impact as the original. More often than not, they think of novel ways of expressing terms, and to be able to so, they draw from their lifetime experience of living in to cultures. As you would know, machines cannot do that.
Machine Translation’s Quality is Not out of the World
Assuming that machines can come close to achieving human translation quality, it is still not clear which version they would emulate. For instance, Give a piece of text to 50 professional human translators and you will get 50 different pieces of translated text. Which one is the best in terms of less number of errors, better reproduction of original text etc? This problem sounds similar to asking a person to pick the best rendition of a sound track when sung by a number of different singers. You may argue that one artist had perfect pitch, while another hit a flat note. Anyway, your choice will be prejudiced. The beauty of human language expression lies in its diversity, which makes everything a bit more complicated. That being said, this diversity is a necessity as well.
In contrast, machine translation can present only limited options with the quality of translation, even though it can deal with complexities to an extent. In other words, their output is normally quite simplistic for the complexities associated with typical translation jobs.
Context Holds the Key
Most languages in this world, like English for example, have specific words that can imply different meanings depending on the context. This is why context matters the most in any form of translation be it human or machine. Peter Gilliver, the lexicographer of the Oxford English Dictionary, stated that just the verb-form of the word “run” has just about 645 different meanings. The question is- can a machine learn all these meanings for a particular word in not only one language, but two?
Humans can use context to good effect to deduce meaning of a word. When we try to figure out the meaning of a word, we think not just of that word in particular, but how that word interacts with the others around it. It takes great ability of professional human translators to handle all the combinations and permutations surrounding the complexity of the word and the other words with which it interacts. Apparently, machines are surely not going to develop the knack that is required to work out different contexts and the different meanings that a single word has in each context. Not anytime soon.