Most Recent Blog Post
Blog Posted:1/20/2014 5:04:00 PM
This week I was looking at an interesting web page done by Charles Bernstein, a pretty well known poet and literary professor. It was a set of experiments, presumably used in poetry classes. What caught my eye eventually were some of the links to text transformation engines, some of which he himself wrote some articles about and experimented with. These included language-to-language translation engines, text replacement engines using a text file and selection of transformation criteria as input and which then outputs replacement text, an engine which transformed text into different dialects- Cockney, Elmer Fudd, etc., a Jabberwocky engine, and so forth. I became interested and also found other engines which generated iambic pentameter, Shakespearean sonnets, and entire poems.
I'm going to discuss a few of these and provide links. Then I want to discuss the use of these engines as tools to generate poetic-like text, which could be used as an aid in the creation of a poem, if not the creation of an entire poem. Hopefully, you will also take a look.
Last, I want to discuss the legitimacy of publishing a poem based on the aid of these engines, which I hope will generate some interesting discussion.
This page is a list of poetic exercises. It contains links to some of the transformation engines as well as an interesting Poem Profiler page. Some of the exercises are pretty cool, too.
Want to translate your poem to a dialect (more or less)? Paste in the text, choose a dialect, and generate. Pretty funny results, and like the other engines to be discussed, will likely require manual tweaking. I might have used the Redneck dialectizer in a recent poem if I had known about this engine, as I spent quite a bit of time going back and tweaking.
Language translator: http://www.bing.com/translator
I used the Bing translator as Babelfish has been replaced by it. Google is also good. I translated a poem of mine into Italian, in which it looked not very poetic at all, then translated the results back into English. Interestingly enough, it translated back pretty well. In fact I liked some of the re-translated wording better than the way I had originally written it.
Text substitution: http://www.spoonbill.org/n+7/
This engine replaces every noun in the text with the 7th one following it in the dictionary. It actually outputs 7th word through 15th word versions. You can get some quite intriguing results with it, as it will generate poetic sounding juxtapositions of unusual words, often however, without semantic context. Those of you who enjoy this kind of tripe will likely love it. See my previous blog “When Wordplay Approaches Jabberwocky” for more grumbling, or better read Mark Twain's “Life on the Mississippi” for his invective.
Language is a Virus: http://www.languageisavirus.com/writing-games.html
This is actually a complete site with all kinds of text manipulation tools. I tried out this one:
Shakespearean Sonnet Generator: http://www.languageisavirus.com/sonnet-generator.html
It just basically pulls random lines from Shakespeare's sonnets and combines them to form a “new” sonnet. The master's jumbles still look very poetic nevertheless, I must say. To really have fun, take the results and run them through one of the other transformation engines, like the N+7 above.
The Pentametron: https://twitter.com/pentametron
The Pentametron watches all the public tweets created in a day. It picks out the ones that happen to be in iambic pentameter. When it finds some of those, it looks for a pair that rhyme, and then it tweets out a couplet. I scrolled through these today and picked out six couplets to form the “poem” below:
The definition of a hustle play
No point in crying over yesterday
I lost myself inside a drunken kiss
Day one and I'm already over this
I never had a valentine before
Whats worth the time and effort anymore?
Don't be a nickel looking for a dime
Its always peanut butter jelly time
I really wanna have a bullet scar
Tuition should include an open bar
Well, changes happen for a reason right?
Don't ask her on a straight tequila night
So what do I mean by these words? I mean a poem created by the combination of human and machine, i.e. a person and one or more text transformation engines, as I did above. At this point in time, there's really not a machine that can generate complete poems even at the level of a good poet, in my opinion. However, I think most of the technology to do so exists, so perhaps there's just no compelling reason to do so, certainly not monetary. But just think: if you have an engine that can spit out iambic pentameter (there is one I found) and it uses as source files the work of Shakespeare, all the classical poets and the best modern ones, uses advanced algorithms to manipulate the text according to rules that you can define on the fly, then it's likely that you can “create” some very interesting poetry. In the meantime however, you can use these tools to help you get past roadblocks when writing, when you're looking for a novel phrase - perhaps just outside the reach of understanding.
Is it cheating? Perhaps, but if so, then it's cheating just slightly more than using rhyming dictionaries like Rhymezone, Wordhippo and the like, in my opinion. The poet, at least for now, still has to assemble the pieces into a coherent (or incoherent) whole and tweak the text.
Now suppose someone used a sophisticated poetry generator with input text from the classics, tweaked it, posted the poem on Poetry Soup, entered it in a contest and placed well. Is there even any way to find out it was done? Probably not. The requirements for reverse engineering such a feat are even tougher than its creation. At most, creating poems this way might begin to show certain patterns, the frequent re-use of the same abstruse words or word combinations. However, this is also true of most poets anyway to some degree, reflecting their style, if you will. You might be caught out only if the poem shows a complete departure from your usual style and vocabulary.
What are your thoughts? Is Cyborg Poetry OK?