Machine versus Human Translation

Mission Statement
Brazilian or Iberian?
Machine Translation
Brazilian Market


Machine translation (MT) is the technology through which computer algorithms render text from one language into another, without any form of human mediation. Its proponents claim to offer a product almost as good as the output of human translators. This is not accurate, as it is shown below.

The use of MT to quickly render Web content into different languages is growing by the day, in response to the pressures to publish content in various languages simultaneously and to offer instant translations on demand to Website visitors. Research from IDC and Forrester shows users are three to four times more likely to buy from a site when they can access its content in their native language. It is therefore not uncommon to find sites that offer free, instant translations into all major business languages, which can be used to convey the core meanings of a Website or an email message, for instance.

On a broad scale, MT has made the Internet more useful to more people. The Internet, in turn, glorifies the communication powers of MT in bridging cultural differences. The two seem to have established an optimistic symbiotic relationship, since the Internet created an emphasis on speedy content turnover.

However, it is our view that MT is not yet ready to replace the more complex work done by human translators. The efforts put into this technology date back to the late 1940s, since when there has been little substantial progress. The subtle variations in context of simple words (which in English can be both a verb and a noun, for example, unlike what happens in Portuguese) can be misinterpreted quite catastrophically by the algorithms that govern machine translation, especially if one does not restrict the context to a narrow subject area.

This technology can only be recommended to assist people in conveying the gist of some simple, specific texts. The detailed adaptation of a source text into another cultural setup, which involves decoding multi-layered strains of meanings from one culture into a totally different code, is an altogether more demanding effort and can only be performed satisfactorily by humans with a sound knowledge of both cultures and a lot of common sense.