Many years ago, back in the days before the Internet and the availability of free on-line translation services, we got a call from a client who told us he had about 300 pages of English he had to have translated into Hebrew.  To determine a rough estimate of the word count, I inquired about the whether the text was single or double spaced, the font and font size and the margins.  The ballpark estimate was roughly 75,000 English words.

When he asked how long the translation would take, I explained that translators generally could be expected to translate about 2,000 to 3,000 words per day.  He quickly did the math and told me that was completely unacceptable—he did not have a month or six weeks of time. I told him it could be accomplished much more quickly if a team of translators were assigned to his project, at an additional fee for expedited delivery.  “I need this by this afternoon…”

He was adamant about his rush situation. “I have this text on a disk.  I can messenger over the disk.”    “And how will that help us?” I asked.  “Well,” he said, “because all you’re going to do is open up the English file and go ‘Save, Save as Hebrew,’ right?”

I burst out laughing and could not stop.  When I tried to control my snickering and giggling, he asked “Are you laughing at me?  Why are you laughing at me?”  I replied: “Because I will be telling this story for a long time.”

Needless to say, this frustrated customer did not get his Hebrew translated by the afternoon that day.  And many years later, I’m sure he feels vindicated by the latest versions of Microsoft Word that do have a translation function built in.  But for anyone familiar with the output of machine translation systems, one of the fundamental aspects of translation quickly becomes apparent, namely that translation “requires the ability to think in two languages.”  And as we all know, machines do not think.

I have to admit that machine translation has made tremendous progress over the years.  Especially Google Translate and similar applications. Rather than relying on software that is programmed to follow rules of grammar and syntax, Google does what Google does best: their algorithms that are so good and so fast, search the Internet for what is statistically most likely to be the best translation.

And while this functionality might be fine for “getting the gist” of a text, and without doubt it will improve, it will still take a person who is intimately knowledgeable of the two languages at hand to sort out and fix such a statistical analysis.  And certain types of translation will resist progress for a long time to come:  I doubt any lawyer, creative copy writer or medical researcher will abandon intelligent humans for an algorithm or a statistical likelihood anytime soon.

So it seems, I still have time to tell this story.

Leave a Comment