Computer Science / Review Essay

Vol. 7, NO. 3 / September 2022

Wacky Jabber

Douglas Hofstadter

Letters to the Editors

In response to “Wacky Jabber


A sweetish suite of machine translations of a pseudo-Swedish paragraph concocted on a whim by this essay’s author.

I  have a pretty long history with Sweden and its language. Actually, until 1961, when I was 16, I’d never given any thought to Sweden at all, but everything shifted on a dime when my Dad shared the 1961 Nobel Prize in Physics. That December, our family flew to Stockholm for the ceremonies and it was unforgettable. Not only were the solemn, yet deeply joyous, festivities a once-in-a-lifetime occurrence, but I was powerfully struck by the classic European beauty of Stockholm in the midst of that romantically dark and snowy Scandinavian winter—such a clean and sophisticated city with its old-fashioned trams, its glittering neon signs, its colorful store windows, its elegant ladies and gentlemen, and, last but not least, its strange, alien language.

When, one day in the Grand Hôtel, which certainly lived up to its name, our family first laid eyes on my Dad’s Nobel diploma, colorfully and exquisitely hand-calligraphed in Swedish, I tried to make some sense of the citation—“För hans banbrytande undersökningar över elektronspridningen mot atomkärnor och därvid gjorda upptäckter rörande nuckleonernas struktur”—but I couldn’t do much with it. Although I knew a weensy bit of German, I didn’t know a word of any of its cousin languages, such as Swedish. And yet I instantly noticed something that looked odd; a tiny little thing that, to my eye, stuck out like a sore thumb. Even if you know zero Swedish, I urge you to try to spot the spot that was an eyesore to me. Hint: it’s just one word toward the end. And by the way, in English, what those italicized words mean is this: “For his pathbreaking investigations of electron-scattering from atomic nuclei, and for his discoveries, made thereby, concerning the structure of nucleons.

Okay, I’ll reveal what made me scratch my head. I knew my Dad’s research was all about atomic nuclei and their constituents—nucleons—and I was pretty sure that the Swedish word for “nucleon” wouldn’t have a “ck” in it, but either just a “k” or a “c.” I didn’t know what the final five letters of “nuckleonernas” meant, but that was irrelevant. So, in a mixture of worry and befuddlement, I pointed out this questionable spelling to my parents, who of course didn’t know what to make of it either. Later I asked some Swedes and found out that my intuition had been right; the calligrapher had simply made a blunder. It should have said “nukleonernas struktur”—literally, “the nucleons’ structure”—with just a “k,” not a “ck.” So, amusingly enough, my Dad’s Nobel diploma has a special distinction, making it very rare, like a postage stamp with a misprint.

Well, that was my first tiny brush with the Swedish language.

In 1965, I was living in London with my parents and sister for a few months. I was feeling restless and, for various reasons, I got the urge to go to Sweden. I had a couple of acquaintances in the old university town of Lund, in southern Sweden, near Malmö, so I contacted them and they urged me to come. And so I did, in early January of 1966. I wound up living in a student dormitory in Lund for about three months and loving it immensely. I made some lifelong friends during that time: Ingemar Börjesson, Birgitta Ohlsson, and Bengt Olle Bengtsson. With them all, I of course spoke English, but from the very start I tried to learn to speak Swedish. I quickly found the language exceedingly alluring, with its curious tonal lilt and its very striking initial consonant-clusters—such as “kv,” “gj,” and “stj”—and of course its unique letters “å” and “Å.” But nearly everyone’s English ran rings around my attempts at Swedish, so it was very hard to get my foot, or even my toe, in the door. One day, however, by luck I met a young Swedish woman on my floor of the dormitory who was very timid about speaking English. She tolerated my stumblings-about in Swedish, so somehow a few conversations with her became min tå in the door, and then they became min fot in the door. After a while I found a few other Swedes who were willing to speak Swedish with me, and from there on I was off and running.

That spring, I moved to Stockholm and lived in various places, mostly amongst students, and became close to the Lawaczeck family, who lived in the Stockholm suburb of Kungsängen. Even after all these years I’m still friends with Yvonne (“Mon”) Lawaczeck Seifert. And with the Lawaczecks I spoke almost exclusively Swedish all the time, and that gave a huge boost to my Swedish. However, I lived in Sweden for only six months total during that year abroad, and in the late summer of 1966 I returned to America to go to graduate school. And with that, my love affair with the Swedish language—or at least my first love affair with it—was over, alas.

It was only in 2016—exactly fifty years later!—that I came back seriously to Sweden and its language. What turned the trick was a tempting invitation I received to participate in a small symposium that June in Stockholm, called “The Limitations of Scientific Knowledge.” My old memories of Stockholm and of course of Swedish were so vivid that I couldn’t resist accepting.

My inviter, Christer Sturmark, turned out to be a very warm and boyish fellow and, on the last evening, he invited a few of us to his home for dinner. It was a lovely occasion and, when Christer found out that I knew some Swedish and loved the language, he presented me with a thick book he had recently written, entitled Upplysning för det tjugoförsta århundradet (Enlightenment for the Twenty-First Century). As he handed it to me, he said, “It’s written in very simple Swedish, so I’m sure you’ll have no trouble reading it.” Well, what Christer meant by “very simple Swedish” was that it wasn’t written in an obscure dialect, wasn’t a trove of medieval poetry, and wasn’t a technical treatise on toads’ skin diseases, or something equally opaque. The truth was, it was written in very flowing, normal, even sophisticated Swedish. But that was just fine with me. On my return flight, I flipped through it and tried to read some of it, and found it both fascinating and fun. As soon as I got home, I sent Christer an email offering to translate his whole book into English. Kind of a crazy idea, but why not? Christer was thrilled and, over the next couple of years, that’s exactly what I did. It was a true labor of love, and soon the plunge I’d taken had fully revived my long-dormant love for Swedish.

The next academic year (2017–2018), I had a sabbatical from Indiana University. My wife Baofen and I decided to spend our year in four marvelous cities in four different countries, immersed in four different languages, one right after the other: the fall in Vienna, Austria (German); the winter in Uppsala, Sweden (Swedish); the spring in Hangzhou, China (Mandarin); and the summer in Geneva, Switzerland (French). It was an amazing year, to say the least, and while in Uppsala I drenched myself once again in Swedish. I did my utmost to speak Swedish with people around me—especially Anders Andersson, Gunilla Borgefors, Ulf Danielsson, Lars Oestreicher, and Christer Sturmark—even though these were all people who spoke nearly perfect English and certainly way, way above my level of Swedish. But somehow, they all very kindly indulged me and I will never forget the many hours I spent speaking Swedish with these sparkling and warm people. It was a second love affair with the Swedish language and a high point in my life.

However, once Baofen and I got home after our amazing international year, I did virtually nothing to improve my Swedish and so of course it started fading. I felt very sad about that, but I had so many other things I was involved in that I just couldn’t keep it up. But a few months ago, for some unclear reason, I felt an urge to read some of the dozen or so books in Swedish that I had acquired in Uppsala. And so I tackled one, and then another, and then another, and another, including two collections of short stories written by Tove Jansson—the Swedish-speaking Finnish author who wrote many famous children’s stories about the “Moomintrolls,” as they are always called in English. Right now I’m in the midst of reading a very stimulating book about genetics and politics written by my old friend from my Lund days, Bengt Olle Bengtsson.

Reading in Swedish affords me much pleasure, but it is not a piece of cake. Each page takes between 6 and 26 minutes, usually around 10, with lots and lots of dictionary lookups—not just for meaning but also for pronunciation, since Swedish is filled to the brim with ambiguities that make it far from perfectly phonetic—and the very careful writing of scads of marginal notes.

As a result of this recent and far from trivial reimmersion in Swedish, my mind has been roiling and boiling in that language—in its sounds, words, idioms, spelling, grammar and so on—and a few days ago, to my great surprise, I found myself randomly making up fake-Swedish words and phrases. This silly and pointless but very playful activity amused me a lot, so on a lark I decided to sit down at my computer and have some fun for a while. In an hour or so, I wound up producing a paragraph that was chock-full of nonsense words that looked and sounded very Swedish—at least to me!—but that, taken all together, meant absolutely nothing.

More explicitly, I made up all the “Swedish” words myself, except for a few pronouns, the conjunctions meaning “and” and “or,” the articles meaning “the” and “a,” the word for “not,” which is “inte,” and a few other very common short words. All the nouns, verbs, adjectives, and adverbs were fictitious—just products of my fantasy. However, I used standard prefixes and suffixes in order to make the past tenses of verbs, the plurals of nouns, agents and actors based on verbs, adverbs from adjectives, and such things. In that sense, it’s very much like Lewis Carroll’s immortal nonsense poem “Jabberwocky”: “’Twas brillig, and the slithy toves…”

I thought my Jabberwockish pseudo-Swedish paragraph was pretty funny and, at first, that was good enough for me. But then it suddenly hit me that I might feed the paragraph to my old frenemies Google Translate and DeepL, just to see what they would do with it. These are two of today’s flagship machine-translation programs and, although—or perhaps because—these programs have improved by leaps and bounds over the past few years, I greatly enjoy discovering and poking fun at their many unpredictable weaknesses. Little did I suspect how enormously much those oh-so-earnest engines’ poker-faced “translations” of my nonsense paragraph would crack me up. When I read their respective outputs, I found myself rolling on the floor. Their wacky jabber was a riot! And then my friend David Moser threw the highly touted Chinese translation engine Baidu into the mix, which added even more humor.

Apparently, what went on behind the scenes—and I hadn’t anticipated this in the least—is that each of these humorless engines “corrected” what it took to be a whole bunch of misspelled Swedish words. It then strung together a set of corresponding real English words, without having the least notion that the text it was ingesting and the text it was spewing out were both totally devoid of meaning, though in vastly different ways. None of these three world-class machine-translation systems had the slightest awareness of being engaged in a marvelously funny GIGO (“garbage in, garbage out”) activity.

In theory, I could spend a few hours converting my droll pseudo-Swedish paragraph into some kind of “equivalently droll” pseudo-English paragraph—doing that would necessarily be a very subjective act of artistic creation. But ars longa vita brevis—and anyway, in the end it would just wind up sounding sort of like Carrollian Jabberwocky. That is to say, a passage made up of lots of nonexistent funny-sounding “English” nouns, verbs, adjectives, and adverbs stitched together into a grammatical-sounding text with the help of short genuine English words such as “the,” “a,” “but,” “and,” “or,” “with,” “I,” “we,” “he,” and standard prefixes like “un-,” “dis-,” “re-,” “con-,” and standard suffixes like “-s,” “-er,” “-ed,” “-ly,” “-ing,” “-tion,” and so on. Just think of “Jabberwocky”—or “Jabbarvåking,” if you prefer—and you’re more or less there.

But to be more concrete, and a bit more verbose, I’ll give you two short examples. Take the following phrase, which occurs near the beginning of my paragraph of pure gibberish: “och sen med de inluppta trämplissorna blybbade det otvickligt.” Here is one way that I might artistically render it in similar pseudo-English gibberish: “and then, with the inlooped tramplezes, it blubbed untwickly.” Pretty silly, eh?

By contrast, here is what the three ultra-rapid language-manipulating engines did with the “Swedish” phrase:

  • Google: “and then with the slivers of wood thrown in, it remained undisturbed.”
  • DeepL: “and then with the drums in, it leaded uncomfortably.”
  • Baidu: “and then with the stuffed wooden bundles it certainly embarrassed.”

Also pretty silly, but unwittingly so. Should I call these ultrarapid machine translation systems “language-mangling engines” instead of “language-manipulating engines”? What would you say, reader?

Okay, now for the second example. My pseudo-Swedish phrase was this: “Det var så att säga hultsamt och multsamt, och även ypperligen gnålfritt.” I might render this in semi-poetic pseudo-English as follows: “It was, so to speak, hultish and multish—indeed, supremishly gnollfree.” Of course I could do the job in a million other ways; that’s just what came to my mind right now. And here again, by contrast, is what the three heroic language-mangling engines did with that same “Swedish” phrase:

  • Google: “It was, so to speak, merry and merry, and also excellently free of whining.”
  • DeepL: “It was, as it were, hulky and overcast, and also exquisitely whine-free.”
  • Baidu: “It was so to speak hulled and mulled, and also excellent whining-free.”

You can decide for yourself which of the three superfast and super-fluent but splendidly point-missing engines did the best job. For me, they all did fabulous jobs, namely, of making me split my sides in rollicking, gleeful, totally off-my-chair-falling guffaws—that is to say, of making me splat my soods in rullicking, glayful, taddily off-my-cheer-failing gafuze. And of course I hope that the same, mutatis mutandis, holds for you, dear reader, whether or not you speak any Swedish.

Without further ado, here’s the full “Swedish” paragraph that I created and fed to the omnivorous engines on August 15, 2022:

Sista mällingen frädde jag mina skvallrutor på en eller två djyvelräckiga drammsniggor, men det knackraddes ogrinligt vålsent för spjulingarna, och sen med de inluppta trämplissorna blybbade det otvickligt. Kvältsmusarna tryckades för tjabbriga i och till de spyrlösa fjöllsidorna, och vi trömmades välmåset med våra innansätsingar. Sen över det ledvist häppligt fliknandet lömnade vi, och flingade den vålfredsskjutliga hjornarens knövboltar. Framligtvist inklågerade jag mig fräsinglöst om att knupa hur de där inblitsna värstningstjuperierna skylle prälla. Vi var yppligt klickforniga över att ingen brusspil eller mänskivel spjällde klotiskt, och att inga förförlingsmässa blyxor näbblades. Det var så att säga hultsamt och multsamt, och även ypperligen gnålfritt, så att jag truppade nästan frukkigt med en gnisskurlig sopperfläcke. Kan man särvist åtvända det där som en tvörlig frösénsstjukare, eller sjöll man brysa en döppel härnäss för att vidstycka sig med rejella äktningsfylla snippflyggor? Det där vulle jag inte bjärkna, inte alls.

Here is Google Translate’s anglicization thereof, created that same day:

In the last mælling, I spread my gossip panes on one or two devilish drum snips, but it was knocked unbelievably late for the children, and then with the slivers of wood thrown in, it remained undisturbed. The voles were pressed too fast into and to the barren fjöllsides, and we were thoroughly beaten with our insinuations. Then over the sadly fortunate flapping, we slapped, and flake the knuckle-bolts of the peace-sick horned. Of course, I complained incessantly about how those blinded worst-case idiots were supposed to act. We were very happy that no noise arrow or man’s disc sputtered, and that no seduction fair lights were snagged. It was, so to speak, merry and merry, and also excellently free of whining, so that I trooped almost frantically with a squeamish soup stain. Can one particularly use that as a sleazy froséen swagger, or should one rush a doppelganger here in order to be swaggered with solid, full-bodied bow ties? I don’t want to spoil that, not at all.

And here is DeepL’s anglicization thereof, also from the same date:

The last time I fed my gossips on one or two devilishly long drams, but it rattled horribly violently for the dampers, and then with the drums in, it leaded uncomfortably. The whelks were pushed too thickly in and to the vomitless sides of the mountains, and we were tired well with our inner seams. Then over the ledvish flicker we lurked, and fling’d the ass-bolts of the fierce-shooting herd. In the meantime, I was busy trying to figure out how to make the worst thieves in the world boast. We were luxuriantly click-happy that no noise arrow or man’s whistle twitched clumsily, and that no seduction fair leadxes beaked. It was, as it were, hulky and overcast, and also exquisitely whine-free, so that I trooped almost fructally with a giggle-crazed sopper. Could one particularly use that as a two-pronged seed stunner, or did one have to break a numb-nosed kidney in order to disperse oneself with a lot of authentic snipe flies? I don’t like that, not at all.

And finally, but certainly not most negligibly, here is Baidu’s anglicization thereof, thanks to David Moser in China. This version joined the collection one day later:

The last thing I said, I cut my gossip on one or two dramma bitches, but there was an unspeakable knock on the vomit, and then with the stuffed wooden bundles it certainly embarrassed. The squirrel mice were pushed too clumsy in and to the puke-free hillsides, and we were filled with our innards. Then over the artificially awkward flipping we lied, and flaked the knuckles of the fearless corner. Mainly locked I’m so excited to find out how those big-time heifers blame me. We were superbly clicked that no noise spike or human disc damped the globe, and that no seductive pencils beaked. It was so to speak hulled and mulled, and also excellent whining-free, so that I marched almost fruitfully with a squeaky soup stain. Can you turn that into a needy frost picker, or you brew a baptismal you’re here to join in with a bunch of real-life snippets? That vulture I didn’t notice, not at all.

In conclusion, I’d like to ponder aloud, so to speak, why it is that I am so amused whenever I read any of these alleged “translations.” Whence all the floor-rolling guffaws?

To begin with, all three machine translations royally miss the point: the input text is made up of nonwords that look and sound like Swedish, but that any Swede would instantly see as non-Swedish and would find quite amusing. To translate the passage faithfully would thus require the creative invention of authentic-looking but non-English words—and moreover, the fake English words would have to be funny non-English words. That’s the key point of the input text.

The fact that all these systems missed the point completely gives me great joy, but that’s not the only reason that I laugh. I laugh also because the passages that they produced are so serious-sounding, and because the three passages are so utterly different in “meaning.”

Of course, none of the three machine-produced paragraphs has any meaning whatsoever, but the systems aren’t aware of that flagrant lack. This is because they have no notion of what meaningfulness and meaninglessness are. They are not thinking while translating; they are just doing very complicated but knee-jerk reflex operations with pieces of text. I say “knee-jerk” for a good reason: they all produce their output paragraphs in a split second—by which I mean a second or two, maybe five at the most—whereas if I were to produce a full pseudo-English text based on the pseudo-Swedish input, it would take me quite a long time, perhaps half an hour or so. I would be roughly 2,000 times slower than these knee-jerk systems. That’s sort of funny, too.

Finally, how can one keep from guffawing if one reads the following “English sentences”?

  • “The voles were pressed too fast into and to the barren fjöllsides, and we were thoroughly beaten with our insinuations. Then over the sadly fortunate flapping, we slapped, and flake the knuckle- bolts of the peace-sick horned.” (Google)
  • “The whelks were pushed too thickly in and to the vomitless sides of the mountains, and we were tired well with our inner seams. Then over the ledvish flicker we lurked, and fling’d the ass-bolts of the fierce-shooting herd.” (DeepL)
  • “The squirrel mice were pushed too clumsy in and to the puke-free hillsides, and we were filled with our innards. Then over the artificially awkward flipping we lied, and flaked the knuckles of the fearless corner. Mainly locked I’m so excited to find out how those big-time heifers blame me.” (Baidu)

“Garbage”, I’m afraid, is almost too kind a word for this kind of output text. Each of the three machine “translations” is an unprecedented example of sheer, total meaninglessness. And yet they were all produced by sober, no-nonsense, deadpan, tone-deaf, and stone-dead programs that have nonetheless been trumpeted in many prestigious and influential publications—such as the New York Times, the Economist, and others—as being astonishingly powerful and supremely accurate translators. It’s as if each program were telling you: “Okay; here’s exactly what the paragraph I’ve just ‘read’ really means. Plus, I produced it in but a split second!” What a farce!

Also, the vast mismatch between the three systems’ output is simply stunning. I guess I laugh so hard because that disparity reveals, like a flash of lightning on a pitch-dark night, the amazing lack of understanding on the part of these highly vaunted and often virtuosic programs. It so vividly brings out their true, zombie-ish nature. And so, these three pieces of crazy, useless, thoughtlessly produced, utterly meaningless garbage masquerading as sense give me immeasurable pleasure.

If you share my feelings of joy and mirth, I’ll be delighted.

If not, well, de gustibus

Endmark

Douglas Hofstadter is College of Arts and Sciences Distinguished Professor of Cognitive Science and Comparative Literature at Indiana University, and a Pulitzer Prize-winning author.


More on Computer Science


Endmark

Copyright © Inference 2024

ISSN #2576–4403