Poems are generally sequences of words that fit some pattern, most often based on rhyming structure or number of syllables. For example, a “haiku” poem is a series of 5 syllables, followed by a series of 7 syllables, followed by a series of 5 syllables. Other types of poems are defined by their rhyme structure – a couplet is a pair of end-rhymed lines. We propose to implement a means of expressing these types of constraints and then automatically solving them to end up with a poem.

An example use of the library is included in demo.scm. You can load this demo from the command line:

$ mit-scheme --load demo.scm

and then interpret an example haiku as follows:

=> (print-poem (interpreter test-haiku))
anemones reside
souffle inoperable
souvenir cleveland

If you'd like to hack on or learn more about the code, all of the code is (theoretically) nicely formatted and reasonably well-commented. Read the source for more details, or email one of the authors -- they'd be happy to hear from you!

Once you have Scheme and a copy of this repository, that's it! Simply (load "interpreter.scm") to have access to all of our code. See usage above for more information.

If you'd like to re-build the source dictionary, you'll need Python and some libraries. Once you've installed Python and pip (a Python dependency manager interface), you can install these libraries like so:

$ cd vocabulary
$ pip install -r requirements.txt

You'll also need to install the rhyme tool, whose modified source is included in the rhyme-0.9 directory. The installation instructions are located in the INSTALL file in that directory.

Once that's done, you can re-run the scraper and re-build the dictionary like this:

$ cd vocabulary
$ make clean
$ make

This will take about 4-5 hours, so unless you have some particular reason to do so we advise against it.