Lost in translation

Errata, by Kevin Young [above, from Jelly Roll, published 2003] is a punch-drunk-in-love-lyric poem that makes intoxicating sense.

Young is giddily postmodern in his focus on the sound and cadence of familiar sweet nothings, to which he applies poetic licence, sampling the lines through distortions, spoonerisms, alliterations, onomatopoeia, inversions of meaning and intent – all by simply nudging letters around, the words still trapped in their proverbial phrasing, their cliches remade into special surprises, or even exposed in their post-truth [“one and only” becomes “won and homely”], structures molten together to reflect the confusion of love.

For the reader, great pleasure comes from resisting the impulse to autocorrect the worn phrases, in the same way skimming text allows the brain to fix misspellings.

I experimented with Machine Learning to take these deconstructive mechanics further…

Processing the text through the constraint-based logic of Google Translate’s AI – I sampled multiple round-trip translations of the piece – handing off from one ‘poetic’ language to another [e.g. English-French-Persian-Urdu-English]. The goal was to try and untangle the hot mess Young had made and hopefully restore the phrases back to their original form and context.

Now, AI uses the statistical repetition of ‘true’ instances to learn – so there was bound to be much confusion from the original ‘false’ input. Further, because translation is also an imprecise process, which needs to intuit and apply complex grammatical and cultural lenses for accuracy.

In a sense, what the Google Translate algorithm [and any repetitive mechanical process] does is the opposite of poetry – it seeks to flatten, standardize and demystify. The programmatic preference is to make all language cliche, to brutally remove any dissonance. In process terms it will flag, remove, correct, apologise – to create the perfect errata.

The result of my experiment was obviously a comedy of erratas – but each layer of output brought its own cadence and meaning – filled with different standards and weights, deviations from the norm and accidental truths.

And what is that, if not the very understanding of the glory of language Young’s original poem reveals?

So, in a way, each translation is fatefully accurate.

Copy and paste the text below [or any piece you love] into translate.google.com, shuck it around in different tongues, build a funhouse bower of table and find your own truth …



Baby, give me just

one more hiss

We must lake it fast


I want to cold you

in my harms

& never get lo

I live you so much

it perts!

Baby, jive me gust

one more bliss

Whisper your

neat nothings in my near

Can we hock each other

one tore mime?

All light wrong?

Baby give me just

one more briss

My won & homely

You wake me meek

in the needs

Mill you larry me?

Baby, hive me just

one more guess

With this sing

I’ll thee shed

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s