Technology is always offering new ways to increase the profitability of every industry. The translation community should not see technology as a threat, but as an opportunity.
It’s very common to hear translators complain about having to work on projects involving post-editing machine translation (PEMT), which tend to pay a lower per-word rate compared to human translation. I have some bad news for those people: we’ve reached a point of no return. This kind of service will become more common as companies—as well as translators—look for ways to cut costs and increase profits. Adapting to this emerging reality is a matter of survival now, and complaining about it won’t negate this situation.
Considering how quickly technology advances, and the fact that most people work for about 40 years, it’s impossible to conceive of a single worker, regardless of field or activity, who will never be affected by technology. This idea sounds frightening, but technological advances represent excellent opportunities that must be seized.
It’s Human Nature to Fear
the Unknown Why would things be any different in the translation industry? I had not even begun working as a translator when computer-assisted translation (CAT) tools were introduced, but more experienced colleagues tell me many opposed them at first. Such a reaction to new technology is hardly unique. Looking back, there even seems to have been some resistance to using typewriters!
It’s really just human nature to fear or resist change. You could even say that this state is a historical truth. For example, such resistance happened back in the 19th century, when automatic power looms were destroyed by Luddites. Despite this extreme reaction, looms were ultimately implemented by the textile industry. Workers who were able to adapt to this new industrial advancement took advantage of the situation and became more successful. So, what does this mean for us?
Will Machine Translation Force Us to Become Proofreaders?
During our careers, we should expect requests to work with machine translation (MT) to increase. That’s why we need to learn as much as possible to turn such requests into opportunities.
Before I graduated with a degree in languages and literature and became a professional translator, my primary activity was inherently a technological one: electronics. This is most likely the reason why I’ve always embraced innovation. I view hardware and software as partners, not as competitors, and certainly not as enemies. I believe far superior results can be achieved when man and machine work as a team, as opposed to two men working alone or two machines working together. A man-machine symbiosis always seems to be superior in both quantity and quality.
Machines take care of repetitive, merely mechanical tasks that don’t require creativity or cognizance—two areas in which humans overpower computers. Despite developments in artificial intelligence, computers still cannot think for themselves or act cognitively. Computers operate based on algorithms, which by definition do not allow them to make the same simple associations that the human brain is able to process. Our immense ability to be creative and make associations, deductions, and decisions allows us to fine-tune our work.
Computers can only do what humans teach them to do in a logical and sequential manner. Therefore, when it comes to MT, translators must tell the computer what it did wrong and how mistakes can be corrected. With this feedback and much-needed human intervention, example-based machine translation (EBMT) and statistical machine translation (SMT) are able to “learn” how to improve the text. Reputable translation agencies that work with MT use feedback from their translators to improve their systems. Google Translate, among others, uses this strategy when it asks you to suggest a better translation for a given sentence. Engineers analyze millions of suggestions. Once they are validated, they are implemented in later software updates.
We would be naive to think that computers will one day have all the correct options available and be able to translate everything perfectly. There are several factors involved in the translation process, mostly of a cognitive nature. This takes us back to the fact that machines are still unable to think cognitively, and will not be able to do so in the foreseeable future. Because of this, there is close to zero chance that MT will prove to be an efficient replacement for human translators when there is a need for high-quality output. What I do see is an opportunity to work side-by-side with MT software to increase our productivity and offer top-quality results.
You’re the Teacher
When I’m not working with PEMT—that is, when I receive a translation request—my workflow includes pre-translation with a client-specific translation memory. After making the most of what the translation memory has to offer, I use rule-based MT software to leverage the glossaries and rules I’ve created. Even though this software may sometimes offer inferior results compared to something Google Translate might suggest, this technique yields two great benefits:
- My client’s material, which is protected by a non-disclosure agreement, is not submitted to any server or website.
- I get to decide how the translation will be done.
The computer is responsible for the mechanical tasks. It applies glossaries, grammatical rules, and style options, over which I have full control. As a result, we have a pre-translated text that will be edited by me before it reaches the client in a ready-to-publish format.
Are the results I achieve using this method always better and quicker than translating it the so-called traditional way? No, not always. I mean, it will be, but not right away, and not without some effort on my part to train the software to do better.
Using a stand-alone MT system installed on your computer requires you to invest some time before it yields rewarding results. First, you must create glossary and fine-tuning rules, as well as rethink your workflow. Additionally, not all types of texts are suitable for software translation. Material that demands more creativity, such as literature or marketing material, and texts that encompass cultural and interpretive aspects, are usually not good candidates for MT. Technical texts are the ones that yield better results.
People who have been working for some time with a specific type of text and, consequently, have specialized in a given field and compiled a well-structured glossary, still need to adapt the way terms have been entered into that glossary. Translators who are used to working with CAT tools may have grown accustomed to adding glossary terms without many restrictions. However, when working with MT, things tend to operate quite differently. For example, in Portuguese nouns must be added to a glossary in their male, singular form. Verbs must be recorded in the infinitive form. Failure to observe these rules will result in a less than useful text, which will demand more from the translator.
Other rules must be set clearly, such as whether the software should use an article before a possessive noun, or if a verb tense will change according to the context. How about the names of buttons on a user interface? Should they be translated or kept in the original language, perhaps with a translation in parenthesis?
You might think, “Well, if I’m going to waste my time micromanaging all that, it’s not worth it!” Well, it is worth it. As anything else in life, you must invest time to achieve good results.
Do you remember a few paragraphs back when I mentioned that computers can only do what humans teach them to do? Well, this means you have to constantly train the software to help it produce results that are acceptable. This requires you to start early and remain vigilant.
It may sound complicated, but what you basically need to do is create a glossary, translate a text, see what needs to be adjusted in the glossary and list of rules, and then translate the text again. With time, you’ll notice fewer errors and you’ll spend less time editing these translations.
In addition to reviewing your material more quickly, you’ll also be developing a new ability as you train the software: PEMT. That’s definitely one more skill you can add to your résumé.
Resistance Is Futile
When it comes to innovation, our industry is passionate about making changes and becoming more competitive. This same passion must be applied to technology. We must take action to make the most of what innovation has to offer. We are a creative species, and we are not easily replaceable. According to John Kelly, senior vice-president of Solutions Portfolio & Research at IBM, “The machines will be more rational and analytic. People will provide judgment, intuition, empathy, a moral compass, and…human creativity.”1
Change is going to happen. More companies will try to find people who are better prepared to deal with innovation. PEMT will certainly be yet another innovation you’ll experience in your career. Instead of bemoaning it, try to get ahead of it!
- Isaacson, Walter. The Innovators (Simon & Schuster, 2014), 486, https://goo.gl/HPWiwN.
William Cassemiro began his career as a translator of technical manuals while still working at Xerox of Brazil. Prior to this, he worked in the electronics industry for more than 20 years. After receiving his degree in English>Portuguese literature, with an emphasis on translation, from the University of São Paulo (USP), he attended courses in translation specialization at USP and the Pontifícia Universidade Católica de São Paulo. He serves on the board of directors of the Brazilian Translators and Interpreters Association. Contact: firstname.lastname@example.org.