In order to complete this project, I had to teach myself Python over the course of a few days and learn to utilizes the tools provided through NLTK. My responsibilities included developing methods to ensure proper syllable count, to parse large batches of txt files into unigrams with (key->value) associations between word frequencies -> count of occurrence in the text. Additionally, I wrote methods to identify parts of speech, and collaborated on a method to weed out parts of speech that sounded grammatically incorrect at the end of haiku lines. I also collaborated on adding AJAX to the haiku and user-feedback form.

Haikuna-Matata generates completely original, correctly formatted haiku based on user input. A user types a word into the console when running the program, and the program will return a 5/7/5 style haiku built around the user's input. Given no user input, the haiku generated will be centered around a randomly generated word or pair of words. Users are then asked to rate the coherence of each line of the haiku. If a line is up-voted, the frequency of those word pairings will increase in the database. If a line is down-voted, the frequency of those word pairings will decrease in the database. Over time, this will train the generator to create more coherent haiku.

All Haiku are defined by the following structure

• Each haiku has 3 lines.
• Each haiku follows the following syllable construction for each line
• 5 syllables for the first line
• 7 syllables for the second line
• 5 syllables for the third line

All Haikus must

• Be centered around a user provided word, or given no user input, centered around a randomly generated word
• Have no line ending in a preposition

Stretch Goals for the machine to learn

• Haikus must make sense.
• All haikus will follow proper grammatical structure.

• Python
• Natural Language Toolkit
• Travis CI
• SQLAlchemy
• PostgreSQL
• jQuery/AJAX
• Flask