misconnections: generative missed connections poetry

I thought it would be interesting to generate poetry using Craigslist Missed Connections postings. I think there is something often sweet and romantic about these posts, and they tend to sound very poetic in their own right. Looking at the site, you get a sense of all these lonely people in New York, looking for one another, yet here they all are, in the same place. Wouldn’t it be interesting to take their texts and recombine them, in effect “reconnecting” all these “missed connections”?

I used several libraries and methods to achieve my results. The first problem to tackle was how to pull live data from the web to use as a source text. Since Craigslist posts can be read as an RSS feed, I found a very useful RSS feed parsing library called Universal Feed Parser to achieve this.

Once I pulled and parsed the data from craigslist, I then used a Markov chain generator class that I got from my professor at ITP, Adam Parrish. After fooling around with the different variables, I decided that using 2grams to generate text was the best way to go, since it was more likely to recombine different texts and give me funny, profound, or interesting results.

Next, I had to come up with some poetic form structure that I felt reflected the content of the text, and the process by which it is generated. A common motif in the missed connections posts are the terms “I was” and “You were,” referring to the chance encounter that didn’t materialize. So, for my poetic from, I created this structure:

4 stanzas that alternate starting with the phrase “I was” and “You were.” These first four stanzas are three lines each in length. The final, fifth stanza starts with a random phrase, and is two lines long.

I thought it would be appropriate to post three of the results back onto Missed Connections:
http://newyork.craigslist.org/mnh/mis/2256880918.html
http://newyork.craigslist.org/mnh/mis/2256876980.html
http://newyork.craigslist.org/mnh/mis/2256871348.html

So far, I’ve had only one response :(
And here are a few more sample results:

Missed Connections 1
I was going to the Danger ZoneGonna take youRight into the Danger ZoneYou'll never say hello to youUntil you get it on the 42nd street A,C,E stop.
I was cute.
Then you got off at myrtle wyckoff.

You were a little older than me and my cocknot sure if it ever happens here but I hope he is not...
You ordered a black hat on,
and a green purse walking from the ferry walking to bowling green trainstation.

I was your waiter,
and I'd love to get a drink sometime :) You: Cute,
asian,

You were sitting across from a very cute brunette wearing black tights or dress pants (not sure) and a dark jacket with dark painted fingernails listening to your iPod.
I am just writing to let you in,
because you were talking with a green purse walking from the ferry walking to bowling green trainstation.

Late afternoon...you were staring at such a beauty..
and you took out a bag of almonds.please come again tonight! you left before i got the nerve to ask you if what language she speak and u keep on smiling.....

Missed Connection 2
I was working asking to use the bathroom.
I replied that the jig is up,
and that the tigress you have back at home sweet home tonight around 11pm,

You were wearing a blue "puffy" jacket,
a beige cashmere hat,
am 6'4" and tan.

I was supposed to knowthat night...I cannot walk away from the big sis.
Why are you single beautiful? so i know its u which side of the steam room or sauna.When you did too.
Email me for more good conversation.

You were taking the M train home last night at a show.
hopefully she didnt pack up and tell me the bank and the love i have for you at the showers and I licked my lips in a sweater dress at Subway on 2nd ave tonight (tuesday)me: the guy you caught me looking at the next-door table.
flesh packed by the pound with broken hearts tightly woundand in the kitchen at the corner with a long brown ponytail and glasses.

I was wearing green basketball jersey and you (cute brunette) were pretty much right next to me at the Bowery back during his Lifted tour;
almost everyone from Interpol (I’ve seen them a lot...and well....I live in NYC...it’s kinda easy);

Missed Connection 3
I was reading into this,
or if you wanted to say how attractive I thought might be yours.
I was too shy to reciprocate.

You were white,
tall and with your friend.
Email me for more good conversation.

I was carrying with me.
- JFyi,
pic is as close as I could find to your Bowie-inspired charm.

You were listening to music on your bathroom window!!! on 37 & 3......i wish i could always use another pal in the groupie world,
you could call use Band-aid’s.
We love the music bizz,

...but I was there with my family...and I could find to your headset,
just remember the hours spent together...dangerous hourslistening to your table of friends,

The line breaks within the stanzas are determined by punctuation, using a regex match for ‘[.,;!?-]‘. Once three lines are reached, a new Markov text is generated for the next stanza. If the Markov text for a stanza is not long enough for three lines, another Markov text is generated until three lines are filled.

I had also experimented with formatting the text using syllable detection, but found that the script I was using for that was not totally accurate, and made it difficult for the script to produce “complete” ideas.

I also had experimented with incorporating automated spelling correction for every word, but found that this was also giving me poor results due to the “unique” type of language generally used by missed connections posters, so I ended up taking that out as well.

Links

This entry was posted in Uncategorized. Bookmark the permalink.

One Response to misconnections: generative missed connections poetry

  1. SIMON KHAN says:

    Hi,

    Very nice thoughts for any missed connections! Thanks for such a nice post!

Leave a Reply

Your email address will not be published. Required fields are marked *

*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>