Translating poetry

You can’t preserve every aspect of a text when translating. A strict word-for-word translation attempts to be faithful to the words but may be ungrammatical in the target language. An idea-for-idea translation is more readable, but still may not convey the style of the original. Translation reminds me of making maps. There have been countless ways to map a round Earth onto a flat piece of paper, each preserving one aspect of the globe at the expense of others. See, for example, the new Equal Earth projection.

Translating poetry is particularly hard because you want to preserve the meaning and the sound at the same time. Translations might choose to preserve the meter but not the rhyme scheme, or vice versa.

Eugene Onegin

Here are a couple poetry translations I find interesting. The first is Douglas Hoffstadter’s translation of the Russian epic Eugene Onegin. The poem has intricate patterns that one could imagine would be of interest to the author of Gödel Escher Bach. Erik Seligman recently described the poem and Hoffstadter’s translation on his podcast. The poem has three levels of structure:

  1. The ends of the lines rhyme in a pattern of ABAB CCDD EFFE GG.
  2. There’s also a FMFM FFMM FMMFMM pattern of “masculine” and “feminine” rhymes, i.e. stress on the ultimate or penultimate syllable.
  3. The poem is written in iambic tetrrameter, with 8 syllables in the lines with masculine rhymes and 9 in the feminine.

Hoffstadter preserves these patterns in his English translation, at the expense of a lot of paraphrasing. Seligman gives the following example, comparing a few lines Nabokov’s more literal translation

Hm, Hm, great reader,
is your entire kin well?
Allow me, you might want perhaps
to learn now from me
what “kinsfolks” means exactly?
Well, here’s what kinsfolks are:

to Hoffstadter’s

Hullo, hulloo, my gentle reader!
And how’re your kinsfolk, old and young?
Pray let me tell you, as your leader,
Some scuttlebutt about our tongue.
What’s “kin”? It’s relatively subtle,
But you’ll tune in if I but scuttle.

The Bible

Another translation I find interesting is a translation of the Bible called The Voice. In addition to the usual team of Greek and Hebrew language scholars, this translation project invited poets to pay attention to the sound of the English text. In particular, their goal was to translate poetic passages so that they sounded like poetry.

In some ways The Voice is reminiscent of the King James translation. The KJV remains popular four centuries later, in part because it simply sounds good. More recent translations have aimed to sound contemporary but not necessarily to sound beautiful.

Here’s an example comparing Psalm 23:4 in the New International Version

Even though I walk
through the darkest valley,
I will fear no evil,
for you are with me;
your rod and your staff,
they comfort me.

and in The Voice

Even in the unending shadows of death’s darkness,
I am not overcome by fear.
Because You are with me in those dark moments,
near with Your protection and guidance,
I am comforted.

The italics indicate words that are not explicitly in the text but are implied, and are added to make the translation flow.

Related posts

Maybe it’s just hard

If someone tells you repeatedly that something isn’t hard, maybe it’s just hard.


A post by Gilad Bracha got me thinking about this. He says

Last time I looked, the Haskell wiki listed 29 tutorials on [monads]. … Could it just be that people just have a hard time understanding monads? If so, what are the prospects of mass adoption?

Monads are not that difficult, but when you have hundreds of tutorials (besides the ones on the Wiki mentioned above) telling you that monads are not hard, maybe they’re hard, at least hard enough for enough people that they’ll never see mass adoption.

I imagine that if something really were much easier if you just look at it the right way, there wouldn’t be that many publications saying so. One or at most a small number of publications would be popular. For example, Mendeleev noticed that the properties of chemical elements were easier to think about if you arranged the elements in a particular tabular form, and everyone agreed, and Mendeleev’s approach won.

The fact that there are hundreds of monad tutorials suggests that many people have independently had an epiphany about monads and wanted to blog about it. I don’t disparage this at all. Many of my blog posts come about this same way: I’ll understand something and write about it while the inspiration is fresh, for my own future reference as well as for the benefit of others.

I also don’t want to disparage monads. They can be very useful. For example, I know of someone who saved his company a lot of money by consolidating a menagerie of inconsistent projects and imposing a monadic framework to make everyone play nicely together. However—and I think this is telling—he was careful to not use the word “monad.” As I’ve written about before, category theory concepts like monads are often most useful behind the scenes.

Regular expressions

This post is the background for my previous post on regular expressions. I didn’t want to say regular expressions aren’t hard; a lot of people clearly do find them hard. Instead, I wanted to focus on why they’re hard, and give my perspective on perceived reasons and more fundamental reasons.


Unicode is similar to regular expressions in that some people find it trivial and some find it maddeningly difficult. Both are right. Unicode can be trivial. The vast majority of pages on the internet use Unicode, specifically UTF-8 encoding. You’re reading Unicode right now. What’s so hard about that?

At its core, Unicode simply assigns numbers to characters, just as ASCII did. In a sense, Unicode is trivial. But it attempts to capture all human writing systems, and writing systems are complicated. As with regular expressions, it’s the peripheral issues that bring the complexity. Unicode is subtle because human language is subtle.

Translating Robert Burns

Last year Adam Roberts had some fun with Finnegans Wake [1], seeing how little he could edit it and turn it into something that sounded like Return of the Jedi. I wrote a blog post where I quantified the difference between the original and the parody using Levenshtein distance, basically how many edits it takes to go from one to the other.

This morning I wanted to post an example of a more likely use of Levenshtien distance. I’m going to look at the final verse of To a Louse by Robert Burns, and compute the distance between the original Scots version and a translation in to more standard English.

Here’s the original:

O wad some Pow’r the giftie gie us
To see oursels as ithers see us!
It wad frae mony a blunder free us,
An’ foolish notion:
What airs in dress an’ gait wad lea’e us,
An’ ev’n devotion!

And here’s the translation:

Oh, would some Power give us the gift
To see ourselves as others see us!
It would from many a blunder free us,
And foolish notion:
What airs in dress and gait would leave us,
And even devotion!

The edit distance between the two versions of the verse is 34. The original has 186 characters, so the translation is about 18% different than the original.

Hirschberg’s sequence alignment algorithm shows how to line up each version with the other.

From Scots to the translation:

    O|| wa|||d some Pow'|r ||||||||the giftie gie us
    To see oursel||s as i|thers see us!
    It wa|||d frae|| mo|ny a blunder free us,
    An'| foolish notion:
    What airs in dress an'| gait wa|||d lea'|e us,
    An'| ev'|n devotion!

And from the translation back to Scots:

    Oh, w|ould some Pow|er give us the gift|||||||||
    To see ourselves as |others see us!
    It w|ould fr||om m|any a blunder free us,
    An|d foolish notion:
    What airs in dress an|d gait w|ould lea|ve us,
    An|d ev|en devotion!

I first heard this poem as a child, and I think about it fairly often. Part of my job as a consultant is to show companies how I as an outsider see their projects. I need the same input, so I turn to advisors to free me from blunders and foolish notions.

Related posts

[1] Of all books I have never read and have no intention of reading, Finnigans Wake is probably the one I’ve referred to the most. It’s so ridiculously difficult to read that it makes good raw material for humorous posts.

Per stirpes and random walks

If an inheritance is to be divided per stirpes, each descendant gets an equal share. If a descendant has died but has living descendants, his or her share is distributed by applying the rule recursively.


For example, suppose a man had two children, Alice and Bob, and stipulates in his will that his estate is to be divided per stirpes. If Alice and Bob are still alive when he dies, his estate is split evenly. Suppose, however, that Alice is still alive but Bob has died, and that Bob has three living children, Carol, David, and Erin. In that case Alice would inherit 1/2 of the estate, and each of Bob’s children would inherit 1/6, i.e. 1/3 of Bob’s 1/2.

State law

In some states, such as Texas, per stirpes is the default when someone dies without a will. Who knows what they do in Nevada? Maybe the descendants play poker for the inheritance. I don’t know. I’m not a lawyer, certainly not a Nevada lawyer.

Random walk

Here’s a random process whose expected value gives the same result as per stirpes.

Convert the inheritance to a jar of pennies, possibly a very large jar. Repeat the following steps until all the pennies are gone.

  1. Take a penny out of the jar and perform a random walk on the family tree representing the descendants of the departed.
  2. When you come to a branch in the tree, choose a branch at random with each branch having equal probability.
  3. When you encounter a living person, give them the penny.

This assumes that you first prune the descendant tree of any lines that have died out. That is, we assume every terminal node of the tree represents a living person.

Why is it necessary to trim the tree? If you ended up at a dead end, couldn’t you just put the penny back and start over? No. Suppose in the example above that Alice and Carol are the only surviving descendants. Then per stirpes says they should receive equal amounts, since Carol inherits all of her father’s share. But if we did the random walk without removing David and Erin, then 1/2 the time we’d give a penny to Alice, 1/6 of the time we’d give it to Carol, and 1/3 of the time we’d start over. Alice would get 75% of the estate.

Related posts

Liminal and subliminal

It occurred to me for the first time this morning that the words liminal and subliminal must be related, just after reading an article by Vicki Boykis that discusses liminal spaces.

I hear the two words in such in different contexts—architecture versus psychology—and hadn’t thought about the connection until now. If I were playing a word association game, my responses would be these.

Q: Liminal?

A: Spaces.

Q: Subliminal?

A: Message.


I checked Etymonline to verify that the two words are indeed cognate. Both come from the Latin word limen for threshold. Something is subliminal if it is below the threshold, typically the threshold of consciousness.


Surely the word subliminal is far more common than liminal. To verify this, I turned to Google’s Ngram Viewer. I’ve included a screenshot below, and you can find the original here.

Ngram of liminal vs subliminal

It’s not surprising that subliminal was a popular term during the career of Sigmund Freud. He published The Interpretation of Dreams in 1899 and died in 1939.

What is surprising, at least to me, is that the word liminal has been gaining popularity and passed subliminal around the turn of the century. I didn’t expect liminal to be anywhere near as common as subliminal.


Google’s Ngram data comes from books. Word frequencies in books can be very different than word frequencies in common speech or other writing as this example shows. I can’t recall ever hearing someone use liminal in conversation. Maybe civil engineers and architects hear it all the time. As I type this, my spell checker puts a red squiggly line under every instance of liminal, showing that the word is not in its default dictionary, though it does recognize subliminal.

Related posts

On this day

This morning as a sort of experiment I decided to look back at all my blog posts written on May 30 each year. There’s nothing special about this date, so I thought it might give an eclectic cross section of things I’ve written about.


Last year on this day I wrote about Calendars and continued fractions, based on a connection between the two topics I found in the book Calendrical Calculations.


Two years ago on this day I wrote about Color theory questions. I’ve been interested in color theory off and on for a while. At one point I thought I might “get to the bottom” of it and figure everything out to my satisfaction. I’ve since decided that color theory is a bottomless well: there’s no getting to the bottom of it. I might pick it back up some day with the more modest goal of learning a little more than I currently know.


I didn’t write a post on May 30 in 2014, 2015, or 2016.


On this day in 2013, I wrote a riff on a quote from Matt Briggs to the effect that there are no outliers, only measurements that don’t fit with your theory.


In 2012 on this day I posted Writing software for someone else. Most of what I’ve read about software development does not make the distinction between writing software for yourself and writing software for someone else, or at least does not emphasize the distinction.

When computer science students become professional programmers, they have to learn empathy. Or at least ideally they learn empathy. They go from completing homework assignments to writing programs that other people will work on and that other people will use. They learn “best practices,” best in this new context.

I made the opposite transition a few months after writing that post when I left MD Anderson Cancer Center to go out on my own. It took a while for me to decide what works best for me, mostly writing software for my own use. Sometimes I deliver software to clients, but more often I deliver reports that require me to write software that the client isn’t directly interested in.


My post for May 30, 2011 was just a quote from Richard Feynman speculating that in the long run, the development of Maxwell’s equations will be seen as the most important event of the 19th century.


In 2010 on this day I posted a quote from Paul Buchheit about the effect of suddenly acquiring wealth. For most people it would not be a good thing.


The post for May 30 in 2009 was called Killing too much of a tumor. You can actually make a tumor more harmful by killing off portions that were suppressing its growth. Reminds me now of how in war you want to leave enough of the enemy’s command in tact that they have the ability to surrender.


Finally, on this day in 2008 I announced that I’d started a web site at reproducibleresearch.org. I later gave the URL to people who had started a similar site with the same name, but ending in .org.

Promoting reproducible research seemed like a somewhat quixotic project at the time, but fortunately it has gained traction since then.


Is there a common theme in these posts? They are all about things that interest me, but that’s necessarily the case since they’re on my blog. One thing that surprises me is that the posts are not particularly mathematical. I would have expected that a quasi-random sample of posts would have turned up more math. But I did write about cancer and software development more when I worked in a cancer center and managed software developers.

Unifiers and Diversifiers

I saw a couple tweets this morning quoting Freeman Dyson’s book Infinite in All Directions.

Unifiers are people whose driving passion is to find general principles which will explain everything. They are happy if they can leave the universe looking a little simpler than they found it.

Diversifiers are people whose passion is to explore details. They are in love with the heterogeneity of nature … They are happy if they leave the universe a little more complicated than they found it.

Presumably these categories correspond to what Freeman elsewhere calls birds and frogs, or what others call hedgehogs and foxes. I imagine everyone takes pleasure in both unification and diversification, though in different proportions. Some are closer to one end of the spectrum than the other.

The scientific heroes presented to children are nearly always unifiers like Newton or Einstein [1]. You don’t see as many books celebrating, for example, a biologist who discovered that what was thought to be one species is really 37 different species. This creates an unrealistic picture of science since not many people discover grand unifying principles, though more find unifying principles on a small scale. I imagine many are discouraged from a career in science because they believe they have to be a unifier / bird / hedgehog, when in fact there are more openings for a diversifier / frog / fox.

Dyson may be taking a subtle swipe at unifiers by saying they want to leave the world looking a little simpler than they found it. There may be an unspoken accusation that unifiers create the illusion of unity by papering over diversity. True and significant unifying theories like general relativity are hard to come by. It’s much easier to come up with unifying theories that are incomplete or trivial.

Related posts

[1] Or at least scientists best known for their unifying work. Newton, for example, wasn’t entirely a unifier, but he’s best known for discovering unifying principles of gravity and motion.

Internet privacy as seen from 1975

Science fiction authors set stories in the future, but they don’t necessarily try to predict the future, and so it’s a little odd to talk about what they “got right.” Getting something right implies they were making a prediction rather than imagining a setting of a story.

However, sometimes SF authors do indeed try to predict the future. This seems to have been at least somewhat the case with John Brunner and his 1975 novel The Shockwave Rider because he cites futurist Alvin Toffler in his acknowledgement.

The Shockwave Rider derives in large part from Alvin Toffler’s stimulating study Future Shock, and in consequence I’m much obliged to him.

In light of Brunner’s hat tip to Toffler, I think it’s fair to talk about what he got right, or possibly what Toffler got right. Here’s a paragraph from the dust jacket that seemed prescient.

Webbed in a continental data-net that year by year draws tighter as more and still more information is fed to it, most people are apathetic, frightened, resigned to what ultimately will be a total abolishment of individual privacy. A whole new reason has been invented for paranoia: it is beyond doubt — whoever your are! — that someone, somewhere, knows something about you that you wanted to keep a secret … and you stand no chance of learning what it is.

Related posts

Impossible to misunderstand

“The goal is not to be possible to understand, but impossible to misunderstand.”

I saw this quote at the beginning of a math book when I was a student and it stuck with me. I would think of it when grading exams. Students often assume it is enough to be possible to understand, possible for an infinitely patient and resourceful reader to reverse engineer the thought process behind a stream of consciousness.

The quote is an aphorism, and so not intended to be taken literally, but I’d like to take the last part literally for a moment. I think the quote would be better advice if it said “unlikely to misunderstand.” This ruins the parallelism and the aesthetics of the quote, but it gets to an important point: trying to be impossible to misunderstand leads to bad writing. It’s appropriate when writing for computers, but not when writing for people.

Trying to please too wide and too critical an audience leads to defensive, colorless writing.

You’ll never use an allusion for fear that someone won’t catch it.

You’ll never use hyperbole for fear that some hyper-literalist will object.

You’ll never leave a qualification implicit for fear that someone will pounce on it.

Social media discourages humor, at least subtle humor. If you say something subtle, you may bring a smile to 10% of your audience, and annoy 0.1%. The former are much less likely to send feedback. And if you have a large enough audience, the feedback of the annoyed 0.1% becomes voluminous.

Much has been said about social media driving us to become partisan and vicious, and certainly that happens. But not enough has been said about an opposite effect that also happens, driving us to become timid and humorless.

Fascination burnout

Here a little dialog from Anathem by Neal Stephenson that I can relate to:

“… I don’t care …”

Asribalt was horrified. “But how can you not be fascinated by—”

“I am fascinated,” I insisted. “That’s the problem. I’m suffering from fascination burnout. Of all the things that are fascinating, I have to choose just one or two.”