Urban Dictionary definition generation using a RNN. Experimentation with text generation with colloquial language, while we wait for our Twitter developer account to be approved. Will probably use a similar model on tweets for an ongoing mystery project ;).
Update: Twitter dev account approved! Something coming soon...
credit to https://www.tensorflow.org/tutorials/text/text_generation for starter code
dataset from: https://www.kaggle.com/therohk/urban-dictionary-words-dataset/data
featured definitions: the following have been generated by the RNN through 10 epochs of training, with a loss of 1.350, down from 4.5455.
bagovern: The feeling yet, unmunctive and smashing country of rap grounds current play. It is an awkward situation that could be painfully 'value'.
storinesence: Something that is in-take out in a wine]
suited problem: Another name for someones handnd cure around. It is very and as a weaking activity of drinking.
sack ready: Birghod, and [chav] of California Maxki Marisachshif Head Marpy Punk band to Do? Occasionally, Mafia fish
Described pillows: 'Holding star?' To have an obscure term, '[performed]' by friends. But it fans must be filled body tighty the time
write.py: python script to parse the data set and convert from csv to one continous text file to use as training dataset. Get rid of 'bad' definitions by only including those with at least 75% positive 'votes'.
text_generation.ipynb: (python notebook, so I can use google colab and their far, far, superior computing power than my puny macbook) Builds and trains model. Then generates some of its own urban dictionary definitions.