Other tools: ee wittgenstein with JanusNode
Empirical reality corresponds a world completely un touched. But that is the description of the following words: what can be arranged in series. Therefore logic and thus the philosophy would fit in to another.
June 10-11, supervised generation from bigrams with heuristic reformatting. Source text: Wittgenstein’s “Tractatus”, Generator: JanusNode 2.05
Using this program was like hooking up with someone else’s girlfriend. It’s a great experience: the program is pretty solid but with a pleasant number of crazy quirks (which is something that both poetry generation tools and girlfriends should have!) But there’s a tinge of guilt: I couldn’t help feeling that I was somehow cheating because I didn’t really know the program as well as I did my own.
Anyway, I’d heard of McPoet before, but I hadn’t realized that it had been renamed JanusNode, ported to windows in the early 00s, and updated early this year at janusNode.com before mysteriously disappearing recently.
Currently the only version available online is a copy of the 2004 Windows version at archive.org. (Google’s cache accessed the latest janusnode.com version as recently as June 4th – if I never get a hold of it and end up missing it by a mere 6 days, I’m gonna be pissed! lol) (UPDATE: it came back! yay!)
It’s not immediately obvious, but clicking on the image of Janus generates poetry as specified in the drop-down menu just below it. These drop-downs apparently select “TextDNA sets” (which seem to be sets of rules?) used to generate the poetry. You can generate your own TextDNA sets by pressing the button and selecting the appropriate “brainfood files.” (awesome!) You can presumably author TextDNA sets and brainfood files yourself, since they are text files. You can paste a bunch of text into the textarea and press “eecummingsfy” or “Dadafy” which will rearrange the text, or “Text Mapping” which will transform it into, say, leetspeak, or ‘olde english’. The whole package came with a 70-page pdf file of imaginatively-written documentation.
But that’s not all. What’s relevant to this current blog: you can create word or character bigram and trigram language models (as in gnoetry, epogees, and mchain) and use them to generate text, which presumably can interact with any of the above.
I just started playing around with it a couple days ago, so this first poem was built using bigrams from a model that came with it. I generated a bunch of text, picked the best 3 lines, and eecummingsified it.
I’m not sure I’ll use it very much more, because I find I gain a lot by having an intimate knowledge of the program that is doing the generation. But it looks like a great tool. Chris Westbury, the developer of JanusNode, is a worthy human being whom I wish to emulate in my own computational poetry endeavours. If I’ve got the right guy, he does neurolinguistics as a TT prof. in Canada. He’s apparently been working on JanusNode as a labor of love (i.e. unfunded!) off and on since the 80s, and in the google cache of janusnode.com he says “I do not intend to definitively cease work on JanusNode in my lifetime” although he only updates it every couple of years. In other words, his approach to McPoet/JanusNode seems pretty close to my own approach to my computational poetry.
Anyway, if you want to try it out, you can hope that janusnode.com comes back (UPDATE: it came back! yay!) and get the Mac or Windows copy, or check out the older archive.org Windows copy (it worked on Windows 7,
I had a bit of a problem on Windows XP though that might be cause I unzipped it wrong.) (UPDATE: yep I unzipped it wrong.) You may eventually turn back to the poetry generation tool that you really love, but you’ll be better for the experience.