Skip to content

a treacherous line

July 26, 2012

from these to emanate
their cries echoed dismay
“We are sure that he cannot reincarnate.”
nervous laughter echoed through the bay.

the other two were Exeter
they felt the double strain and tug;
he will be there next to her,
the treacherous line smug.

her very choice:
new jersey.
she’ll read joyce
on the anniversary.

half choked with sewer gas
none save the rats will pass.

 

sonnet generated by prosaic and extended by hand from 200,000 phrases gleaned from project gutenberg.

Advertisements
5 Comments leave one →
  1. July 29, 2012 6:10 pm

    Great ending couplet. Keep these sonnets coming.

  2. nathanielksmith permalink
    July 31, 2012 1:30 pm

    thanks! i’ve found that the couplet ends up doubled most of the time, so i end up writing an original line to go with. the next feature planned for prosaic is an optional second-pass which de-dupes the poem.

  3. August 1, 2012 12:45 am

    Very interesting stuff.

    First stanza – I like lines 2-4. Line 1 is a little weak; I like the word “emanate” but it doesn’t really follow from the previous words in any interesting way I can see.

    Second stanza – again, lines 2-4 are strong, but I have problems reading the first line in a creative way. (though “Exeter”, which is necessary for the rhyme scheme, is probably salvageable.)

    The third stanza’s brevity gives it the cuteness of a limerick. It’s very effective as a stanza, but I’m not sure how in tone it fits in with the rest of the poem.

    The final couplet is great, as eRoGK7 points out.

    I haven’t had time to look at prosaic much, so I’m not sure what kind of authoring approaches would work with it. Maybe what you want is something that generates a first draft, then generates several candidates that end in a given word? For example, if you decide to change the first line of this sonnet, you could get several candidates that end in “emanate.” (or, of course, you could just human-author it, but I find machines provide brilliant juxtapositions that just need to be sought out by humans.)

    Another challenge is coming up a poem made of stanzas that cohere in some interesting way. But that could be as simple as just generating a number of stanzas and manually clustering those that are similar in tone.

    Anway, an interesting direction, thanks for sharing.

  4. nathanielksmith permalink
    August 1, 2012 10:13 am

    thanks for the thorough review: it’s very helpful as i’m still determining the shape of my collaboration with prosaic (as a tool and algorithm).

    I’m trying to keep the human-authorship to a minimum and further juxtaposition supplied by the software is the goal: i like your idea of using one poem as a “seed” for another. that certainly would have been helpful here as I really liked a few of the lines and had to work to salvage the rest (commuting them humanly into something that seemed to fit).

    currently you cannot supply a rhyme sound to use at any point (rhyme sounds are chosen at-random in the beginning and then used throughout the generation) but I think that is a feature I would like to add (ie, instead of just given a symbol to represent a rhyme slot actually provide a word to rhyme against).

    As a sidenote, I also intend to have prosaic report on which sources it used in a given run so I can properly attribute each poem…

  5. August 2, 2012 1:09 am

    I’m trying to keep the human-authorship to a minimum and further juxtaposition supplied by the software is the goal: i like your idea of using one poem as a “seed” for another. that certainly would have been helpful here as I really liked a few of the lines and had to work to salvage the rest (commuting them humanly into something that seemed to fit).

    Yeah, deciding what role the human plays in authoring is a big part of it. At one extreme, you’re making a spellchecker-like tool to assist principally human generation. At the other extreme, the program generates full poems that are presented without further editing. Back when I started, I decided not to worry too much about whether the human input devalued the effort too much. (i.e. my goal was to make interesting generators or interesting poems, rather than try to make some kind of fully-automatic poetry generator). But clearly that’s not the only approach, and there is value in trying to have a computer do as much as possible.

    I think one of the big problems in fully-automated generation is adding a discourse structure, i.e. having all the stanzas cohere in some way. Knight et al use n-grams at the line level: i.e. the selection of a word is partially decided based on the words that occurred in the previous LINE as well as previous n words. (if I remember correctly…) An approach I wanted to try at some point was lexical chaining: in each line, have at least one word that was related to a given set of words: i.e. either be a synonym, antonym, hyponymn, holonymn, etc, using something like WordNet. (I generated some poems that I manually verified contained at least one word in a given chain i.e. view senses death and delusion.) There are also various discourse models based on essays such as rhetorical structure theory; an analogue of these could probably be developed for poems.

    A lot of different possible approaches. All depends on what you find interesting.

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 )

Google+ photo

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

Connecting to %s

%d bloggers like this: