Web Gadget

The rise of the machines

05 Aug 2013 by
| Filed in Web World
The rise of the machines

Cisco estimates that global cloud traffic will grow 45% annually until 2016, with translation services growing at around 15% to 20% per year. This means that many new machine translators must enter the industry each year to handle the content, says IAN HENDERSON, CTO of translation and localisation company, Rubric.


According to Ray Kurzweil, a futurist known for his predictions about artificial intelligence, machines will match human intelligence and perform feats including human-quality translations by 2029. 
 
Current happenings also suggest a strong role for non-human translation, with machine translation (MT) advancing rapidly of late. Three simultaneous-translation devices have been announced since June 2012,
 including one by Microsoft that renders live audio translations from the spoken word, in the tones and inflexions of the speaker.

Perfect is hard
But perfecting a translation machine remains one of the
 toughest challenges in artificial intelligence. For decades, computer scientists tried using a rules-based approach — teaching machine translation systems the linguistic rules of two languages and giving it the necessary dictionaries. 

Then researchers at companies like Google began to favour a statistical approach. By feeding the computer thousands or millions of passages and their human-generated translations, it could make accurate guesses about translating new texts. 

Google has been reported as saying it doesn’t want to replace human translation, but merely aid in broadening access to the vast volumes of information on the Internet. 

But even if Google wanted to take over the world, it would be hard. Machine translation tools simply cannot take into account the purpose, real-world context or style of any utterance.

Humans need machines
Machine translation (MT) – not to be confused with computer-aided (human) translation (CAT) – involves the use of software to translate text or speech from one natural language to another. It is particularly effective in contexts where standard or formulaic language is used, such as legal or government documents.

Machines need humans
MT is perceived to be highly efficient at translating high volumes, but anybody who works in machine translation will appreciate the difficulty of separating the human element out of the process. 
 
In many cases, human intervention is needed to edit the source text before, and the translation after, the translation process. Humans are also needed to train the computer to deal with a specific topic and terminology. For this to be cost-effective, a very large volume of words must be processed by the system. Recognising this, some companies use MT initially, and then employ human editors to iron out problems, perceiving this to be a smart and fast way of combining man and machine. 
 
The best of both worlds
A third alternative, computer-aided translation, combines the best of both worlds. CAT technology leverages the speed of machines as well as the human knack for understanding context and nuance, to deliver large-scale translation project outcomes that compare favourably with the best human translators at a speed that holds its own with machine translation services like Google.

CAT tools and techniques acknowledge the mutual reliance of machines and humans, and build a high degree of integration into the process to deliver high-quality translation at scale, quickly and with fewer resources than human translation.
 
When to use MT, CAT or human translation
That being said, there are instances in which sacrificing accuracy in favour of speed or vice versa is entirely adequate. The following scenarios can help with deciding between MT, human translation, combined efforts or CAT.
 

· A consumer trying to find a product on a Chinese site – MT. Accuracy is not likely to be in question with the use of commodity product names.

· Getting an email in the wrong language – MT. Subsequent communications will settle whatever ambiguity there may be.

· A copywriter seeking an understanding of research data – MT followed by human translation to speed up the ramping-up process and refine the report with 100% accuracy.

· A technology company seeking a website translation (e.g. a search engine) – CAT. CAT tools aid in retention of the translation memory of companies. Consider also the cultural sensitivities around certain brand names.

· House style – depending on volumes, CAT. Most companies prefer certain tonalities over others and have certain linguistic conventions or preferred nomenclature. Machines can’t help.

· Context – depending on volumes, CAT. When you’re talking about ‘step’, MT may have difficulty deciding whether it means “step along the way”, “rung in a step ladder” or “dance type”.

· Recurring translation work with incremental updates – CAT. Even a high volume of work should be done by humans upfront, where accuracy is important. The benefit of this becomes clear later, when updates to the text are simply translated as integral to the stored translation in a CAT database.

· One-off small documents – human. The overhead cost in setting up CAT or MT is unlikely to be recouped when dealing with one-off small document translations.


Be guided by circumstances
Circumstances will dictate which form of translation is best for you. Companies that let their translation efforts be guided by their situation will make the best use of their resources and manage risk better.

 

* Follow Gadget on Twitter on @GadgetZA


 email this to a friend
 printer friendly version

 

Share the love

Digg This Article Delicious This Article Stumble Upon This Article Reddit This Article Technorati This Article Tweet This Article Google Bookmarks This Article Bookmark This Article Multi This Article Facebook This Article Link This Article Google Plus This Article

 

Comments on 'The rise of the machines'

Leave your comment




For security reasons, please enter the letters in the image to the left
in the box below; please note that the text is case sensitive.