Generative Art Info

Contact me if you find a bug or have an idea or a suggestion: E.g. if you find a badly formatted poem or would like to see the works of another poet included.


This app generates a haiku (or a tanka) by randomly picking three (or five) lines with 575(77) syllables from well-known poets whose works are in the public domain.

Uses jQuery, PHP, and a MySQL database.

The lines are sorted by syllable count beforehand, using Python, NLTK, CMUdict (Carnegie Mellon University Pronouncing Dictionary), and syllabes. These methods of counting syllables aren’t very accurate, however, and many improvements are necessary.

Read about the poetry corpora that were used.

Poem Hopper

This app consists of two parts: First, it displays a random poem from a poet you can specify. Then, once a poem is displayed, you can click on any word and the app displays another poem that includes this word.

Uses jQuery, PHP, and a MySQL database.

Read about the poetry corpora that were used.

Synaesthetic Metaphor Generator

With this app you can generate a synæsthetic or surreal metaphor or simile either completely at random or by restricting one or more of the word classes to a specific sensory category.

Uses jQuery and (nlp-)compromise.

This app is very much a work in progress, firstly because there are many general improvements that need to be made, secondly because there are many more ways of creating metaphors that need to be added, and thirdly because the word lists need to be ameliorated by removing some words, by recategorizing some words, and by adding a lot more words.

Poetry Corpora

Poetry corpora are from the TextGrid Repository, Project Gutenberg / GutenTag , and Œuvres. The corpora are processed with Python, Bash, Pandoc, Sed, and Perl and other tools.

Unfortunately, these corpora aren’t completely regular and use rather different TEI XML structures, sometimes within one and the same file, which is why I’ve regularized them somewhat using Python and Regex, but some comments, notes, dates, line numbers, bad formatting, unnecessary special characters, etc. still made it through and will need to be removed in the future so that only pure poetry is left.