January 21, 2016

Winners of the Algorithmia Shorties Contest Announced

We're excited to announce the winners of the first-ever Algorithmia Shorties contest. Inspired by NaNoGenMo, the Algorithmia Shorties were designed to help programmers of all skill levels learn more about natural language processing. We challenged developers to get started with NLP by creating computer-generated works of short story fiction using only algorithms. Entries were judged by their originality, readability, and creative use of the Algorithmia platform.

Grand Prize Winner: Big Dummy's Leaves of Spam

Algorithmia Shorties - Big Dummy's Leave of Spam

Our winner of the Algorithmia Shorties is Big Dummy's Leaves of Spam by Skwak. We loved this entry because it creatively mashed up the classic poem Leaves of Grass by Walt Whitman, and Big Dummy's Guide to the Internet, a sort of layperson's guide to the internet, with "select works" from her Gmail Spam folder. The result is a highly readable, and often hilarious poem about the philosophy of life, the Internet, humanity, and scams.

Skwak used Sentence SplitGenerate Paragraph From Trigram, and Generate Trigram Frequencies from Algorithmia. In addition, she created an Express app and hosted it on Heroku. Here Github repo can be found here.

Read Big Dummy's Leaves of Spam here. 

Honorable Mention: n+7

Our first honorable mention is n+7 by fabianekc. It was inspired by a class in Mathematics & Literature he took in 2001. He learned about the "n+7 method," which is where you replace each noun in a text with the noun in a dictionary seven places after the original. For example, "a man meets a woman" is transformed to "a mandrake meets a wonder" using the Langenscheidt dictionary. The n+7 entry features the first four sections of Flatland by E. Abbott translated as a corpus. What we loved about this entry was that fabianekc created an algorithm on Algorithmia to accomplish the n+7 method. The algorithm takes either a link to a text file for processing, a pre-processed text file, or plain text, and replaces each noun with the noun in the dictionary seven places away. You can also change the dictionary, the offset (how many places after the original noun to use), and the start and end of the text to begin the replacement. Check out the Github repo here.

Read n+7 here.

Honorable Mention: Arrested Development Episode

Our second honorable mention goes to a computer-generated Arrested Development episode, created by peap and jennaplusplus. The duo generated trigrams using the scripts from the first three seasons of Arrested Development for all characters that had more than 10 lines overall. This created an eye-popping 71 trigram files! To create the faux episode script, they started the episode off with the Narrator speaking, and then randomly selected which characters would speak based on the size of their trigram file, ensuring no character spoke twice in a row, which resulted in every character having 1-5 lines every time they spoke. Check out their Github repo here.

Read Arrested Development Episode here.

How The Algorithmia Shorties Were Generated

For algorithmically-generated short stories, users start by finding a corpus of text available in the public domain through Feedbooks or Project Gutenberg. Other interesting corpora users could use are TV scripts, software license agreements, personal journals, public speeches, or Wikipedia articles. Users then generate trigrams, which are a collection of all the three-word sequences in the corpus. With trigrams generated, the last step is to reassemble the trigrams into sentences using Random Text From Trigrams, or Generate Paragraph From Trigrams.

Want to learn more? For a detailed walkthrough, check out our trigram tutorial here.


Matt Kiser

Matt Kiser

More Posts

Here's 50,000 credits
on us.

Algorithmia AI Cloud is built to scale. You write the code and compose the workflow. We take care of the rest.

Sign Up