Public Works and Government Services Canada
Symbol of the Government of Canada

Institutional Links

 
Search TERMIUM Plus®

Is dictation outmoded?

André Guyon
(Language Update, Volume 10, Number 2, Summer 2013)

When I was a translation student, we were advised to try dictating our translations instead of typing them. At the time, a lab test had shown that dictation resulted in a significant increase in productivity.1

Thirty-odd years later, young people tell me they’re taught that dictation is outmoded, that it’s for dinosaurs who have trouble using new technologies.

That’s what I call using a huge cannon to kill the goose that lays the golden eggs. I don’t understand why some people decided that translation tools must inevitably replace dictation instead of complementing it. Many language professionals say that dictation is not for everyone, even though they’ve never tried it themselves.

Yet almost all of us have the gift of speech. The fact is that some people do not express themselves very well in writing, but do when speaking. Did you know that roughly seven years of learning and practice are required to reach the necessary level of reading and writing to be able to understand sufficiently well and write decently?

Not only spelling, syntax and grammar, but also calligraphy must be learned in order to write well. Those who are less gifted at drawing, as I am, spend hours practising how to form letters properly.

Speaking, however, is less complicated. Children don’t usually take courses to learn how to speak. They often speak very well before even knowing the alphabet. They can look at picture books and learn the names of things depicted in them.

As a comprehensive system of communication, speaking came long before writing. Speaking comes naturally to us. Children don’t need formal teaching in a school for seven or eight years. Learning how to write on your own, however, is a lot less common.

That being said, even though writing came after speaking, mediums for producing and keeping text (pen and paper) were accessible to the masses well before mediums for recording speech (audio and video). Knowledge was first disseminated in writing, even though talking is our most natural means of communication.2 With the democratization of the Internet, chatting and social media have made a style of writing that resembles casual conversation widespread.

Dictation for efficiency

Despite all this, we should never forget that anyone required to produce a text can save a ton of time with dictation. In my opinion, the pen and keyboard preceded speech recording and speech-to-text devices only for technical reasons. However, now that speech recording and speech recognition are becoming more widespread, these techniques are an obvious part of the evolution of how we produce text. After all, speaking to produce a text would essentially be coming full circle, and rightly so, since the purpose of writing is often to reproduce speech.

The question we should be asking ourselves is, should we use speech recognition software or continue to make use of the expertise of data entry specialists?

Have you guessed what my answer is? To support my point of view, I’ll compare machine translation and speech recognition with respect to producing text.

Software developers create amazing products that do perform incredible tasks, but still somewhat imperfect tasks when it comes to producing text. Choosing between four or five options is something a machine can do very well, but processing language, which offers endless possibilities, is another matter altogether.

For example, teaching my speech recognition software to distinguish between your and you’re is extremely difficult, even impossible. It still sometimes mixes up the two sounds, even though I make sure to pronounce them differently.

I dictated this article using speech recognition software.3 Then I made corrections to it with my keyboard and had it reread by a colleague.

This process saves me time. But I’d still rather work with a dictation entry specialist than use software. Unlike a machine, the specialist would never write "your" where "you’re" is required, even if I mispronounced it slightly. The specialist also wouldn’t write "I like" if I had said "I’d like." The software, however….

The fact remains that a lot of time is saved by combining dictation and speech recognition—a lot more than by typing. However, the time saved would be significantly lower if I used speech recognition to translate, because I usually work by writing over top of the source language in the target language. A specialized editing program like those used in translation tool environments that process sentences one by one in a window with two columns (one for the source language, the other—blank—for the target language) is therefore better suited to speech recognition.

It’s also possible that speech recognition functions specially adapted to such an environment would produce much better results. For example, with my current software, if I want to replace "your" with "you’re," I have to tell my software I want to correct "your," wait, specify that I’m referring to the 18th occurrence, and then choose one of the many options. It’s such a tedious process. However, if the software was restricted to the sentence or paragraph I’m working on, the correction could be made almost instantaneously.

Lastly, because speech recognition software will always make some mistakes, why not have data entry specialists review texts produced automatically? I think this profession needs to evolve and, like others, make use of new technology, such as speech recognition.

Data entry specialists are already experts at formatting in a wide range of software. They could just as easily become experts in correcting mistakes made by speech recognition software and even provide us with suggestions on how to improve that software. This neglected profession would thus be given new life.

The future

Olaf-Michael Stefanov, who was in charge of providing IT services to librarians and language services at the United Nations, initiated an experiment in which data entry specialists could transcribe a text using the traditional method or dictate it to software.

The experiment showed some potential, even though it raised concerns, but it was abandoned at an early stage. Why not use a system that combines human talent and software? Be that as it may, those who tell future language professionals that dictation is outmoded are preventing them from seeing the future!

Notes and references

  • Back to the note1 Details on the experiment can be found in Lise Laroque-Divirgilio’s article "La traduction au magnétophone," Meta, Vol. 26, No. 4, 1981, http://www.erudit.org/revue/meta/1981/v26/n4/002573ar.pdf.
  • Back to the note2 However, my kids and their friends sometimes make me wonder whether text messages are now replacing talking.
  • Back to the note3 I have used Dragon software since 1995, partly for fun and out of interest in new technologies, partly because my joints react to different temperatures and partly because it’s faster than using a keyboard.