Rise of the Machines
Common Sense Advisory released a report on “Trends in Machine Translation” today. Their key findings are:
- Customer experience and engagement mandate more translation. Common Sense Advisory’s research has repeatedly shown that both consumers and business buyers prefer reading information in their own language. (Nothing earth-shattering in either of those statements.)
- The vast amount of content makes MT inevitable. (Ditto – how many of you have relied on Babelfish or Google Translate in a pinch?)
- Over the next few years, every organization’s content strategy will rely on some type of machine translation. Displacing each percentage point of the current spending on human-delivered translation could double the number of words delivered in a foreign language through editing of the output by professionals. For post-edited MT, quality is already equivalent to human translation in many cases. Taking the next step to high-quality trained MT at microcents a word could increase translated content by orders of magnitude.
I disagree with Common Sense Advisory’s third assertion regarding the quality of machine translation. From my observation of the marketplace, MT is just not there yet, regardless of the algorithm used to translate words and phrases. There is still a stunted, choppy awkwardness to more complex phrases and sentences that make it not quite ready for prime-time. Personally, I would rather challenge myself to read a website in the original language rather than stumble through text that feels like a person thumbed through their grammar school English textbook to rewrite.
SpeakLike’s translation engine runs on people power. From experience, the human touch in applying style and elegance to the written word still reigns supreme. In fact, the system has some checks to flag text that appears machine-translated, providing some assurance that the customer did not pay for an important document to be run through some webpage and spit out a mediocre translation courtesy of AI. This is the key value added from SpeakLike. It will be interesting to see what the future holds, but for now, machines do not reign supreme.
The Perils of MT “Chinglish” in The New York Times
The New York Times highlighted some hilarious examples of “Chinglish” recently.
Op-ed in The New York Times about Google Translate
There’s a very interesting op-ed in The New York Times about the strengths and shortcomings of Google Translate. The gist of it is that while Google Translate is very powerful, certain language pairs remain deficient and the system still has a lot of trouble with anything that might be termed “literary translation.” A quote: “For works that are truly original — and therefore worth translating — statistical machine translation hasn’t got a hope.”
SpeakLike featured in The Economist Magazine
This week, The Economist published a great article about how machine translation and crowdsourcing are changing the way people communicate on the internet. Read about SpeakLike in the article.
Monitor: The many voices of the web | The Economist
The internet: New combinations of human and computer translation are making web pages available in foreign languages.
If Machine Translation Were Perfect
Yesterday, we saw some great articles about Google Translate and improvements in machine translation. The articles provide a good perspective on the current state of machine translation and its importance in everyday life to get the gist something in another language.
Unlike a few years ago, we now have a choice in translation: machine or human. So it’s important to understand which one to choose. The New York Times’s version of the story highlights benefits and the limits of machine translation:
“Creating a translation machine has long been seen as one of the toughest challenges in artificial intelligence … Automated translation systems are far from perfect, and even Google’s will not put human translators out of a job anytime soon. Experts say it is exceedingly difficult for a computer to break a sentence into parts, then translate and reassemble them.” Miguel Helft, New York Times
One of our users learned a lesson about this choice recently. In a hurry to respond to a customer’s support question, the user tried machine translation for a quick understanding of the topic. He then rushed a reply (using SpeakLike’s email translation tool). A few minutes later the SpeakLike translation of the original question came back (via SpeakLike’s Translate Text page). He realized that the question was different than how the machine translated it. The question was so different that his answer was embarrassing. In today’s world, and perhaps forever, business communications across languages need a human perspective to convey appropriate meaning.
What would the world be like if machine translation were perfect? It would be integrated into our daily communications and software, in the background. For example:
- Websites and blogs would automatically be posted in the languages of our customers shortly after we added new content or posted updates, consistent with our marketing message.
- You could compose an email in your language and have it delivered in another language, confident about what it said.
- You could chat across languages with (near) real-time translation that reflected what you were actually saying.
- Customer support systems would use support agents who speak the primary language of the company to support customers who speak other languages.
This is what SpeakLike is about — enabling us to use our current methods of communication and still do business in any language.