Some of these cookies are essential to the operation of the site,
while others help to improve your experience by providing insights into how the site is being used.
Do we have the real scope over Machine Translation?
The following article explains in a very objective way what will happen or what are the expectations regarding this software. Knowing the history is as important as known why research on MT ended on 1966 and the effort to develop software based on that research continues. The importance of the topic has been exposed in several occasions, but at the end nothing clear has been settled. MT has been directly mainly, to scientific and technical content and it is clear that Literature is not a field for this software…
Machine translation over fifty years by W. John Hutchins University of East Anglia
“The history of machine translation is described from its beginnings in the 1940s to the present day. In the earliest years, efforts were concentrated either on developing immediately useful systems, however crude in their translation quality, or on fundamental research for high quality translation systems. After the ALPAC report in 1966, which virtually ended MT research in the US for more than a decade, research focused on the development of systems requiring human assistance for producing translations of technical documentation, on translation tools for direct use by translators themselves, and, in recent years, on systems for translating email, Web pages and other Internet documentation, where poor quality is acceptable in the interest of rapid results.”
When it comes to learning a foreign language, we tend to think that children are the most adept. But that may not be the case – and there are added benefits to starting as an adult.
“Not everything goes downhill with age,” says Antonella Sorace, a professor of developmental linguistics and director of the Bilingualism Matters Centre at the University of Edinburgh.
She gives the example of what is known as ‘explicit learning’: studying a language in a classroom with a teacher explaining the rules. “Young children are very bad at explicit learning, because they don’t have the cognitive control and the attention and memory capabilities,” Sorace says. “Adults are much better at that. So that can be something that improves with age.”
A study by researchers in Israel found, for example, that adults were better at grasping an artificial language rule and applying it to new words in a lab setting. The scientists compared three separate groups: 8-year-olds, 12-year-olds, and young adults. The adults scored higher than both younger groups, and the 12-year-olds also did better than the younger children.
This chimed with the results of a long-term study of almost 2,000 Catalan-Spanish bilingual learners of English: the late starters acquired the new language faster than the younger starters.
I’d like to share my latest post with you. It’s the first of a series of posts on Legal Translation and Plain Language. This one, in particular, focuses on how legal translators can benefit from interpretation theory, mainly textualism. It’s called “Legal Translation and Plain Language: What Translators Can Learn from Judges.” Here’s a little extract:
Textualism, intentionalism, and purposivism are not the only interpretation theories out there, but they are arguably the most prominent. Applied to translation, textualism is analogous to asking ourselves what a competent source language speaker would understand if aware of the fact that the text at hand is a legal text. Intentionalism is analogous to asking ourselves what the drafter of the source text subjectively intended to convey. And purposivism is analogous to asking ourselves what the reasonable objective meaning of the source text is, even if that means digressing from the plain meaning of the text.
None of these involve literality, yet applied methodologically to translation, all can result in high degrees of fidelity to source and, more importantly, help translators construct objective criteria when interpreting a text. However, unlike judges, translators don’t always have access to sufficient background information to identify the intention of the drafter, be it subjective or objective. And, unlike the drafter, translators don’t have the power to decide whether the target text should be written in plain language or not. That is, of course, up to the drafter or client.
What translators can do, instead, is adopt a more methodological approach to interpretation—one that, to some extent, emulates how judges approach legal texts and helps unravel the meaning that needs to be captured from the source text and conveyed in the target text. That, however, involves being able to extract the plain meaning of a text, which in turn, also involves understanding what plain language is and how it can be helpful.
NPR has been celebrating National Poetry Month in the US on Twitter, using the hashtag #NPRPoetry. A recent post posed the question of whether poetry can be translated or if “Poetry is what gets lost in translation.” Poet and translator Aaron Coleman translates one of the poetry submissions and discusses his method.
“I approach translation even knowing that it can’t quite be what it is in the original language,” he says.
The language lapses that inhibit an ideal interpretation can ultimately be “a creative, productive failure,” he adds. “Maybe it can open up a new way for us to see what can happen in English and what can happen in Spanish, for me, or whatever the original language is.”
Instead, translation can be transformation. “I think we all want to have translation work as a process of reproduction, but it’s really a process of transformation,” Coleman says.
Just as services like Google Maps have made it easier to get around, at home or abroad, some advances in translation tech are making it easier to travel in places where one may not know the language.
Right now this translation tech is still in its infancy and primarily eases logistical complexities. With Google Translate, you can point your camera at a street sign to verify it says “Downtown This Way” and not “There Be Dragons.” The same app can quickly translate a menu—if not always perfectly, then well enough to be sure that you’re ordering chicken.
The article goes on to consider how far this technology might improve, up to a future where we’re all wearing earpieces and hearing or being simultaneously interpreted for in our conversations. Sound familiar?
“What is it like to live day-to-day as a translator? What are the worries and the stresses, the pleasures and the reliefs? How does a translator get by, and where do her projects fit into the rest of her life? In this new year-long feature, translator Emma Ramdan gives us some answers by keeping an open diary about a year her life.”
Interpreting in 2018 is becoming progressively more of an audiovisual experience in remote encounters, than the face-to-face meetings of the past. As such, in my opinion, there is one issue that interpreters of the 21st century need now more than ever: EMPATHY, the ability to understand and SHARE the feelings of another human being.
As a trained actor from my days of youth, I believe that many of the techniques that are used by actors should be used by remote interpreters; as conduits of the thoughts of another being, those thoughts never exist in a vacuum. Thoughts are intimately related not only to our culture and the patterns of our society but also to our feelings, for thoughts control feelings (and feelings influence thoughts).
Good actors make us suspend disbelief and see THROUGH them the character that they are portraying. We see those “other” human beings (they portray) in all their strength and frailty because the actors are able to get themselves “out of the way” and BECOME a true conduit of the thoughts and feelings of the character they portray.
Good actors, therefore, achieve selfless results (i.e., we see “another” instead of the actor) by developing total EMPATHY for their character. So too, it is my belief, that we, as interpreters, are conduits for the expression of another’s words and feelings. In that sense, it is too little to ask that we “simply” convey words. We MUST convey the words in total accuracy, but we must ALSO convey the thoughts and feelings that are attached to those words. EMPATHY allows us to do so, or at least to try our best. It is this human-ness that will indeed separate us from the likes of bilingual Siris!
Empathy is, at its simplest, awareness of the feelings and emotions of other people. It is a key element of Emotional Intelligence, the link between self and others, because it is how we as individuals understand what others are experiencing as if we were feeling it ourselves…
Three Types of Empathy
Psychologists have identified three types of empathy: cognitive empathy, emotional empathy and compassionate empathy.
Cognitive empathy is understanding someone’s thoughts and emotions, in a very rational, rather than emotional sense.
Emotional empathy is also known as emotional contagion, and is ‘catching’ someone else’s feelings, so that you literally feel them too.
Compassionate empathy is understanding someone’s feelings, and taking appropriate action to help.
So, how do we develop EMPATHY? There are many techniques and exercises. I found some very interesting by Martha Beck, appropriately called The Empathy Workout:
EXERCISE 1: LEARNING TO LISTEN
…start with conversation. Once a day, ask a friend, “How are you?” in a way that says you mean it. If they give you a stock answer (“Fine”), repeat the question: “No, really. How are you?” You’ll soon realize that if your purpose is solely to understand, rather than to advise or protect, you can work a kind of magic: In the warmth of genuine caring, people open up like flowers….
EXERCISE 2: REVERSE ENGINEERING
Some mechanical engineers spend their time disassembling machines to see how they were originally put together. You can use a similar technique to develop empathy, by working backward from the observable effects of emotion to the emotion itself. Think of someone you’d like to understand…Remember a recent interaction… Now imitate, as closely as you can, the physical posture, facial expression, exact words, and vocal inflection they used during that encounter. Notice what emotions arise within you. What you feel will probably be very close to whatever the other person was going through…
EXERCISE 3: SHAPE-SHIFTING
In folklore, shape-shifters are beings with the ability to become anyone or anything. As a child, I was fascinated by this concept and used to pretend that I could instantaneously switch places with other people, animals, even inanimate objects… I recommend you try it, soon. See that strange man in the orange polyester suit putting 37 packets of sweetener into his extra-grande mochaccino with soy milk? What if— zap!—you suddenly switched bodies with him? What would it be like to wear that suit, that face, that physique? What impulse would lead to sugaring a cup of coffee like that, let alone drinking it?
EXERCISE 4: METTA-TATION
World-class empathizers…conduct a daily regimen of metta, or lovingkindness, meditation. This involves focusing all of one’s attention on a certain individual and offering loving wishes to that person with each breath you take, for several minutes at a time. Classic metta practice starts with your own sweet self. For five minutes, with each breath, offer yourself kind thoughts… Then switch the focus of your kind thoughts onto a friend or family member. When you feel a sense of emotional union with that person, target someone you barely know….
55% of freelancers spend 3 hours a week on their online marketing efforts
51% of respondents considered marketing too time-consuming, and 41% felt marketing was too costly
83% are investing financially in online marketing of some sort
72% say they are spending less than but up to 100 USD a month in marketing (those who spend more than that report earning more)
The average survey respondent had reached their income goal within two years of starting out
The survey sample were some 2,000 US freelancers of all types, so it is reasonable to expect those numbers to be somewhat different if we narrow it down to translators and interpreters, expand the sample to other countries, or both.
One number in particular that caught my eye was the monthly investment in marketing. 100 USD a month sounded pretty steep to me, but maybe I’m wrong. 1,200 USD in freelancer marketing a year. Do you spend that much on your marketing? If so, drop me a line, I’d be very interested in hearing about it and if you find it to be a good investment.
If you are already a paying ProZ.com member, you are spending between 12 and 18 dollars a month on marketing through your membership, though you get all the rest of the tools and opportunities available along with it. It’s a kind of marketing that is easy to do, what we’d call passive marketing.
Be an ant, not a grasshopper
For some kinds of work, sending CVs, applications, emails, calling or meeting potential clients, printing business cards or flyers, posting ads, and all of that active marketing, can be effective. Many freelance translators and interpreters find that kind of marketing tiring, frustrating, and also expensive, both in terms of money and time. You’d rather be translating or interpreting, right?
You may have to rush to do active marketing if you suddenly find yourself short on clients or workload. This tends to happen when a freelancer has no kind of marketing in place while they are fully-booked, a bit like the grasshopper who watched the ant stock up for winter, unworried during the summer because food was plentiful, and then sorely unprepared for the winter.
Passive marketing is your ant stocking up for winter. It can help save you from the unexpected, even though work might be plentiful now. And sometimes it’s a gateway to new opportunities that can pop up and replace what you’ve got going on with something even better.
Where is your shop window?
As I said, passive marketing is easier to do, if you do it right, and the time/monetary investment is quite different too. It basically consists of opening up a brightly-lit shop window (your online presence) on a bustling street. Many people walk by, window shopping, but if your shop has the right goods (your services, expertise, samples, things that make you stand out), shoppers will pop in to look and talk to you. Some will be interested in buying now, some will simply make a note of your shop for when they do need what you have to offer.
Where is that bustling street, though? Well, ProZ.com is one of them. You should have a professional online presence in any serious work-related venue for language professionals (a profile on LinkedIn, for example). But since ProZ.com is the busiest street when it comes to searching for and finding language professionals, if you are not figuring there as prominently as possible, you are definitely missing out on client contact. So that ProZ.com membership, roughly the cost of a new pair of shoes per year, is all you need to keep your shop window on the busiest street in the industry.
Check your directory ranking in your top language pair and area of expertise. What page of the results are you on? How many pages of results will your ideal client browse through to get to you? They say, “The best place to hide a dead body is on page 2 of Google search results.” Directory results work in a similar way. Chances are, by the time a client has gone a few pages in, they’ve already found the people they are looking for.
Don’t waste my time
Now, when I say “online presence” I don’t mean having a profile registered on a place and having the bare minimum of information filled out there. Nowadays, if I’m looking for a service/service provider online, I don’t even look twice at people who have not put some time investment into presenting their services. No picture? No thanks. No real name? See ya. No details about the services you offer or why I should choose you? Don’t waste my time! This is where the time investment comes in. It’s mostly an up-front investment. Put in the time to craft that presentation, then go do whatever else you want, and let it go to work for you in the background.
Now think about your two biggest clients…
I’ve got all the clients I can handle right now, no need, you might say. OK! But how many times would you try to go back to a shop that was closed every time you went there?
Now, think about your two biggest clients. Would you be in trouble if tomorrow, through no fault of your own, you lost those two clients? If so, why not put your shop window out there, and occasionally field an inquiry from an interested potential client? The worst that can happen is that you’ll make some new contacts while you’re working, and heaven forbid your fully-booked status should change, you’ll have some good leads to work with.
What’s in your shop window?
Now go over to your ProZ.com profile. At the top of your profile you will see a link to “Force visitor view”. Click on that. What you see is what any visitor to your profile will see when they are evaluating working with you. Put yourself in the shoes, or eyes, of your ideal client. Does what you see there look professional, attractive, keep your interest, “sell” you on the idea of contacting this person with a work offer? Does it speak to that person’s strengths, what makes them different from the competition?
By the way, if you are looking for ways to build, update, or fine tune your online presentation, many of the same principles of decorating a real shop window apply! Thinking about it this way may also help get your creative juices going. If you need some inspiration, you can find some pointers here: https://www.appearhere.co.uk/inspire/blog/how-to-dress-your-shop-window
In this episode we talk about a new tool for freelance translators. I am all for efficiency and organization, but I am also quite lazy, and have struggled keeping track of my projects, number of words and how valuable different projects have been for me, how long they took etc. But now there is a tool that is super easy to use, that does all this for me, and much more. I interview the co-founder and co-creator of the tool Caroline Bries.
Important things mentioned in this episode:
LSP.expert as a project management tool for freelance translators
All the functions in LSP.expert – quoting, job tracking, expenses, income, reports, invoicing, outsourcing, timer and much more
Security and support for LSP.expert
Useful links mentioned in this episode:
Review of LSP.Expert by Silver Tongue Translations
LSP.Experts Facebook page
How LSP.Expert revolutionized my business – Review on The Open Mic
I just got notified, that PayPal is changing its Tems & Conditions on May 25th 2018.
You should give it 5 minutes and check the conditions concerning your country of residence and also which countries are the ones, you get the most transactions from. Mind, that there’s a huge difference if payments are based on bank accounts or credit cards.
I e.g. just realized, that it makes a big financially difference, that my account is registered in Germany, not Austria, and have to make the necessary changes soon.
How do you handle your PayPal fees anyway? Are you simply accepting that loss of money or do you add the costs in your invoices, so the client has to pay them?
Business is looking good in the language sector. CSA Research’s business confidence survey of the CEOs of the largest language service providers found 2017 to be a growth year, and respondents optimistically entered 2018. Sector revenue and language output continue to rise as the content and code that power economies are becoming more global. Our annual survey will give a more complete picture of the market as we collect and analyze the data.
This optimism plays out against a backdrop of concerns about the future of language services, on both the demand and supply side of the market. Buyers worry about the need to process ever-growing content volumes into more language pairs – but with relatively stable budgets. Meanwhile, they must deal with their management’s expectations that Amazon and Google Translate will take care of that pesky language problem once and for all – and with less complexity and at a lower cost.
On the supply side, LSPs express fundamental anxiety about the sustainability of their business models. We hear concerns across all tiers of the language service market:
Automation and procurement specialists marginalize small providers. These LSPs wonder where they fit in the market and how they’ll survive. They worry about sales, staffing, and the need for more – and competition with – increasingly powerful technology. Will their translation work be replaced by a bunch of Amazon servers? Will their project management value be replaced by bots? Further, they must also contend with commoditizing forces beyond their control such as distant procurement functions created when their clients are acquired by global behemoths.
Market forces squeeze mid-sized companies from both ends. Further up the value chain, medium-sized companies face niche specialists on one side, generalist multi-language vendors on the other, and those same procurement challenges. These mid-tier firms strategize about how they can scale up more quickly and compete against the economies of scale that the largest LSPs bring to bear. They plan growth both organically and by strategic liaisons or acquisitions.
The largest providers scramble for scale. At the top of the pyramid, the largest LSPs position themselves to get even bigger, scale to their clients’ fondest dreams for global content, strive for organic growth, and engage in pitched battles to buy a shrinking pool of mid-sized acquisitions. They’re also investing more in building their own technology as they climb to the higher stages of LSP Metrix maturity.
In conversation after conversation, we witness debates about technology. On one side, advocates pitch the importance of advanced technology to growth and scalability. On the other, naysayers anxiously await the swarms of AI-driven language bots and machine learning that will surely exterminate their companies. Their concern extends beyond machine translation to project management − will simple rule-based automation and deep machine learning conspire to eliminate the last humans on the language shop floor? CSA Research characterizes this angst as techno-phobia − or maybe more precisely, phobAI or its homophone FOBAI, the fear of being AI’d out of existence.
But we contend that FOBAI is a symptom of a bigger problem – mistaken identity. LSPs concerned about having automation wipe them out think they’re in the translation business. They’re not. Language service providers are business process outsourcers (BPOs), traditionally tasked with the job of rendering one language into another but now on the cusp of a much broader role managing more of the content assets for their digitizing clients. While conventional agencies still quibble about how many pennies they charge to translate a word, their buyers are far more interested in bigger content issues tied to their digital transformation. These clients are bringing all their information online, optimizing it, adapting and adopting technology to manage it, and making those assets available across the enterprise. This digitization has forced them to re-think internal systems and processes, pry open databases and content management systems, and educate their staff and suppliers in this new model.
How will their clients’ digitization strategies transform the LSPs supplying them?
We’re in a localization and globalization market now where more words are translated every day through machine translation than what was translated in the entire human language corpus in the past.
Not only does such a massive amount of machine translation radically change the role of human translators, it also creates a whole new range of issues that impact the translation and globalization paradigm itself.
And one of the most important issues is ethics.
In an era when entire translations or at least substantial parts of them are often done by machine instead of by professional translators, what does it mean to provide “services” from an ethical perspective as far as translators and LSPs are concerned?
In this week’s episode of Globally Speaking, our hosts Renato Beninatto and M.W. Stevens discuss this very important issue that affects everyone involved in the language industry—both providers and buyers of translation services alike.
Major topics include:
What needs to be disclosed to buyers and what doesn’t?
Are language professionals now selling a product or a service?
When are translators in breach of a client contract by using machine translation, and when are they not?
Why machine translation is unlikely to ever replace the need for professionally trained translators.
How do LSPs charge for projects in which machine translation plays a major role?
Jack Welde helps companies make more money by speaking their customers’ language — literally. Welde is the Co-Founder and CEO of Smartling, a disruptive translation services company that uses a combination of human and machine translation to help companies enter new markets faster.
Welde says in the interview that consumers are 75 percent more likely to convert when they are being sold to in their native language — even if they are comfortable with the language they’re reading. Smartling measures the accuracy of translations with data, as well as the translations’ effectiveness in reaching new customers.
In this episode we are talking about what methods work to find translation clients in 2018. With me I have a translation company owner and translator, Sherif Abuzid, who is sharing his best tips for finding clients. These are suggestions based on his experience. Pick the ones that work for your situation, depending on experience and preference, but also depending on your location.
Important things mentioned in this episode:
Change in how we find and contact clients during these last 10 years
What resources for finding clients we should focus on
How we should contact new clients
Differences in marketing if you are a newer translator vs a more experienced one
By and large, those of us in the localization industry are multi-cultural and multilingual, traveling the world, celebrating the diversity and power of language. We are interested in languages, culture, and language technology, and how they influence and shape our lives. Many of us also cross over from linguistics and language to the technical side of things.
Jost Zetzsche, a localization professional with nearly 20 years of experience, is a thought leader in these areas. He put his expertise into his latest book, Translation Matters. Comprised of 81 essays collected over the past 15 years, Jost’s book describes a world of translation where technology changes rapidly, but where the translator remains the central figure, ever-savvier in using the tools of the trade.
I had the chance to sit down with him and talk about these ideas.
Viju: One very hot topic on the minds of translators today is Machine Translation (MT).
Jost: In part, my book deals with the identity of translators at a time when MT has clearly become important. It’s critical as translators to define who we are and, on the basis of that self-perception, understand our role in the world and our role in relation to things like machine translation. That’s what the book is about to a certain extent.
The way we translators usually approach MT is either we reject it and say, “I don’t want to deal with machine translation,” or we say, “Okay, then I guess I have to do what I’m told to do with machine translation.” And that’s typically post-editing. And while in some situations, post-editing is the right choice, more often than not there are better ways of dealing with machine translation than to not use it at all or only post-edit. For example, translators can greatly benefit from data that is being suggested by machine translation engines without actually “post-editing” complete segments.
Viju: When I think about MT, I also think about another industry hot topic: AI. What are your thoughts on that?
Jost: I just read an article that talked about artificial intelligence, and it said that translators are kind of the canary in the coal mine. If translators become extinct, then we have truly reached a point of no return, where everything has completely changed, where everything has been turned upside down, and where artificial intelligence has essentially taken over.
Translators have a very secure job for a long time to come. If their job becomes insecure, if artificial intelligence, machine translation, is truly able to take over from translators, society will have changed so much that we will not recognize it anymore. Then, essentially, we’re all out of a job. We’ll have to redefine what it means to be productive, what it means to be a human being, what it means to work in the industry, etc.
At that time, there will be what Ray Kurzweil and others are calling this moment of singularity where we don’t need to work in the kinds of professions that we work in today. But, I think translators will be among the last to go.
What translators do is very close to what it means to be human: to be able to communicate in a complex and multi-layered manner. That’s something that computers can’t do unless we reach that point of singularity.
But don’t misunderstand me—it’s not as if we’re not impacted. But our job is NOT being taken away by it.
Viju: What happens to a translator who’s not inclined to learn about the whole range of tools available? Is there space for such a translator anymore?
Jost: Absolutely. You don’t have to be a tech geek to be a translator. My background is the opposite of technical. But I find joy in making technology work for me in a way that is productive and thinking of new ways of making it even more productive. And I think that’s something that translators should have. So that’s not geeky—it’s just looking at what’s out there and finding the right tools that work for you. I don’t think translators need to know all the tools. They just have to have an idea of what is out there and then make intelligent decisions on which tools to use.
One Sunday, at one of our weekly salsa sessions, my friend Frank brought along a Danish guest. I knew Frank spoke Danish well, since his mother was Danish, and he, as a child, had lived in Denmark. As for his friend, her English was fluent, as is standard for Scandinavians. However, to my surprise, during the evening’s chitchat it emerged that the two friends habitually exchanged emails using Google Translate. Frank would write a message in English, then run it through Google Translate to produce a new text in Danish; conversely, she would write a message in Danish, then let Google Translate anglicize it. How odd! Why would two intelligent people, each of whom spoke the other’s language well, do this? My own experiences with machine-translation software had always led me to be highly skeptical about it. But my skepticism was clearly not shared by these two. Indeed, many thoughtful people are quite enamored of translation programs, finding little to criticize in them. This baffles me.
As a language lover and an impassioned translator, as a cognitive scientist and a lifelong admirer of the human mind’s subtlety, I have followed the attempts to mechanize translation for decades. When I first got interested in the subject, in the mid-1970s, I ran across a letter written in 1947 by the mathematician Warren Weaver, an early machine-translation advocate, to Norbert Wiener, a key figure in cybernetics, in which Weaver made this curious claim, today quite famous:
When I look at an article in Russian, I say, “This is really written in English, but it has been coded in some strange symbols. I will now proceed to decode.”
Some years later he offered a different viewpoint: “No reasonable person thinks that a machine translation can ever achieve elegance and style. Pushkin need not shudder.” Whew! Having devoted one unforgettably intense year of my life to translating Alexander Pushkin’s sparkling novel in verse Eugene Onegin into my native tongue (that is, having radically reworked that great Russian work into an English-language novel in verse), I find this remark of Weaver’s far more congenial than his earlier remark, which reveals a strangely simplistic view of language. Nonetheless, his 1947 view of translation-as-decoding became a credo that has long driven the field of machine translation.
Since those days, “translation engines” have gradually improved, and recently the use of so-called “deep neural nets” has even suggested to some observers (see “The Great AI Awakening” by Gideon Lewis-Kraus in The New York Times Magazine, and “Machine Translation: Beyond Babel” by Lane Greene in The Economist) that human translators may be an endangered species. In this scenario, human translators would become, within a few years, mere quality controllers and glitch fixers, rather than producers of fresh new text.
Such a development would cause a soul-shattering upheaval in my mental life. Although I fully understand the fascination of trying to get machines to translate well, I am not in the least eager to see human translators replaced by inanimate machines. Indeed, the idea frightens and revolts me. To my mind, translation is an incredibly subtle art that draws constantly on one’s many years of experience in life, and on one’s creative imagination. If, some “fine” day, human translators were to become relics of the past, my respect for the human mind would be profoundly shaken, and the shock would leave me reeling with terrible confusion and immense, permanent sadness.
Each time I read an article claiming that the guild of human translators will soon be forced to bow down before the terrible swift sword of some new technology, I feel the need to check the claims out myself, partly out of a sense of terror that this nightmare just might be around the corner, more hopefully out of a desire to reassure myself that it’s not just around the corner, and finally, out of my longstanding belief that it’s important to combat exaggerated claims about artificial intelligence. And so, after reading about how the old idea of artificial neural networks, recently adopted by a branch of Google called Google Brain, and now enhanced by “deep learning,” has resulted in a new kind of software that has allegedly revolutionized machine translation, I decided I had to check out the latest incarnation of Google Translate. Was it a game changer, as Deep Blue and AlphaGo were for the venerable games of chess and Go?
I learned that although the older version of Google Translate can handle a very large repertoire of languages, its new deep-learning incarnation at the time worked for just nine languages. (It’s now expanded to 96.)* Accordingly, I limited my explorations to English, French, German, and Chinese.
Before showing my findings, though, I should point out that an ambiguity in the adjective “deep” is being exploited here. When one hears that Google bought a company called DeepMind whose products have “deep neural networks” enhanced by “deep learning,” one cannot help taking the word “deep” to mean “profound,” and thus “powerful,” “insightful,” “wise.” And yet, the meaning of “deep” in this context comes simply from the fact that these neural networks have more layers (12, say) than do older networks, which might have only two or three. But does that sort of depth imply that whatever such a network does must be profound? Hardly. This is verbal spinmeistery.
I am very wary of Google Translate, especially given all the hype surrounding it. But despite my distaste, I recognize some astonishing facts about this bête noire of mine. It is accessible for free to anyone on earth, and will convert text in any of roughly 100 languages into text in any of the others. That is humbling. If I am proud to call myself “pi-lingual” (meaning the sum of all my fractional languages is a bit over 3, which is my lighthearted way of answering the question “How many languages do you speak?”), then how much prouder should Google Translate be, since it could call itself “bai-lingual” (“bai” being Mandarin for 100). To a mere pilingual, bailingualism is most impressive. Moreover, if I copy and paste a page of text in Language A into Google Translate, only moments will elapse before I get back a page filled with words in Language B. And this is happening all the time on screens all over the planet, in dozens of languages.
The practical utility of Google Translate and similar technologies is undeniable, and probably it’s a good thing overall, but there is still something deeply lacking in the approach, which is conveyed by a single word: understanding. Machine translation has never focused on understanding language. Instead, the field has always tried to “decode”—to get away without worrying about what understanding and meaning are. Could it in fact be that understanding isn’t needed in order to translate well? Could an entity, human or machine, do high-quality translation without paying attention to what language is all about? To shed some light on this question, I turn now to the experiments I made.
I began my explorations very humbly, using the following short remark, which, in a human mind, evokes a clear scenario:
In their house, everything comes in pairs. There’s his car and her car, his towels and her towels, and his library and hers.
The translation challenge seems straightforward, but in French (and other Romance languages), the words for “his” and “her” don’t agree in gender with the possessor, but with the item possessed. So here’s what Google Translate gave me:
Dans leur maison, tout vient en paires. Il y a sa voiture et sa voiture, ses serviettes et ses serviettes, sa bibliothèque et les siennes.
The program fell into my trap, not realizing, as any human reader would, that I was describing a couple, stressing that for each item he had, she had a similar one. For example, the deep-learning engine used the word “sa” for both “his car” and “her car,” so you can’t tell anything about either car-owner’s gender. Likewise, it used the genderless plural “ses” both for “his towels” and “her towels,” and in the last case of the two libraries, his and hers, it got thrown by the final “s” in “hers” and somehow decided that that “s” represented a plural (“les siennes”). Google Translate’s French sentence missed the whole point.
Next I translated the challenge phrase into French myself, in a way that did preserve the intended meaning. Here’s my French version:
Chez eux, ils ont tout en double. Il y a sa voiture à elle et sa voiture à lui, ses serviettes à elle et ses serviettes à lui, sa bibliothèque à elle et sa bibliothèque à lui.
The phrase “sa voiture à elle” spells out the idea “her car,” and similarly, “sa voiture à lui” can only be heard as meaning “his car.” At this point, I figured it would be trivial for Google Translate to carry my French translation back into English and get the English right on the money, but I was dead wrong. Here’s what it gave me:
At home, they have everything in double. There is his own car and his own car, his own towels and his own towels, his own library and his own library.
What?! Even with the input sentence screaming out the owners’ genders as loudly as possible, the translating machine ignored the screams and made everything masculine. Why did it throw the sentence’s most crucial information away?
We humans know all sorts of things about couples, houses, personal possessions, pride, rivalry, jealousy, privacy, and many other intangibles that lead to such quirks as a married couple having towels embroidered “his” and “hers.” Google Translate isn’t familiar with such situations. Google Translate isn’t familiar with situations, period. It’s familiar solely with strings composed of words composed of letters. It’s all about ultrarapid processing of pieces of text, not about thinking or imagining or remembering or understanding. It doesn’t even know that words stand for things. Let me hasten to say that a computer program certainly could, in principle, know what language is for, and could have ideas and memories and experiences, and could put them to use, but that’s not what Google Translate was designed to do. Such an ambition wasn’t even on its designers’ radar screens.
Well, I chuckled at these poor shows, relieved to see that we aren’t, after all, so close to replacing human translators by automata. But I still felt I should check the engine out more closely. After all, one swallow does not thirst quench.
Indeed, what about this freshly coined phrase “One swallow does not thirst quench” (alluding, of course, to “One swallow does not a summer make”)? I couldn’t resist trying it out; here’s what Google Translate flipped back at me: “Une hirondelle n’aspire pas la soif.” This is a grammatical French sentence, but it’s pretty hard to fathom. First it names a certain bird (“une hirondelle”—a swallow), then it says this bird is notinhaling or not sucking (“n’aspire pas”), and finally reveals that the neither-inhaled-nor-sucked item is thirst (“la soif”). Clearly Google Translate didn’t catch my meaning; it merely came out with a heap of bull. “Il sortait simplement avec un tas de taureau.” “He just went out with a pile of bulls.” “Il vient de sortir avec un tas de taureaux.” Please pardon my French—or rather, Google Translate’s pseudo-French.
We translators may be on the verge of dealing with more math surrounding our beloved words. Until now, the number that mattered most has been word count, because that is how we are paid. And fuzzy matches, of course—that magic algorithm that calculates how much partial help (and payment) we get.
We consider a segment short, medium, or long based on its word count. However, machine translation (MT) offers different ideas about different lengths; even neural MT with all the widespread praise for fluency.
For segments of one or two words, there is little to no fluency. So is that where neural MT goes out the window? Maybe, or maybe not. We have to find out. Either way, all of sudden, segment word counts have become metrics indicative of whether MT will translate fluently. Food for thought: how many words is still short? What is medium? What is long?
Another number to consider is edit distance. It may be coming to a translation productivity tool near you. That is because we will increasingly want to know how much MT is helping us do our job, and the answer to that (or a part of the answer) is how many changes we make to the suggestions we receive.
If the MT is really awesome in a 30-word sentence, we smile, make two changes, and the edit distance is low. If the MT requires reordering the words and fixing brands that were inaccurately translated, then we are putting a lot more effort and the edit distance will reflect that.
As much as I love edit distance, it is just a comparison between one initial “image” of the sentence and the final “image” that we created. But then there is adaptive technology, that keeps changing the suggestions as we move. Working with adaptive technology, there isn’t really an initial “image” or “initial full sentence translation.” So, it is complicated.
Let’s explore some other numbers: what is the average number of words per sentence in our text?
If it is high—let’s say 16—it means longer and more fluent sentences, and maybe we should expect neural MT to lend us a hand with it. But if we are translating software strings or mobile content, life isn’t like that, is it? Our average word count per sentence is pretty low; our strings are mostly short (except for messages). So we may not get as much help from MT.
Maybe the average number of words per sentence would be an interesting number to take into account when we get a new project?
All of that said, these numbers would be auxiliary metrics. The word count that we use today may be looking at the end of its life. With adaptive technology and varying qualities of MT suggestions, it will become very difficult to associate all the possible variations of those suggestions to overall word count.
After all, fuzzy matches are based on translation memories, on a predetermined percentage of similarity to an entire segment previously translated. How would we use such a strict concept for all the different parameters when working with MT and adaptive technology?
Today, most of the time, the combination of MT and fuzzy matches constitute a system that outputs some mixture of “fuzzy grid for 75 / 85% and above” and “discounts for MT suggestions below that threshold.” End-clients and agencies calculate edit distance to get an informed view of whether the MT is still helping us (or if it is helping too much). And that is the system: “fuzzy with some MT discount.” Edit distance is also great to detect patterns that will improve MT output: if we’re making lots of changes, the MT may need some improvement.
But with adaptive technology and who knows what other forms of AI-powered augmented translation, how are we going to fit the fuzzy grid into this? We won’t. Translators will be paid based on time spent working. So everybody will change how they work and everybody will win. Dear translators, start thinking how much your hour is worth. It is about time.