Trumpitter

Trump Tweet Generator using Markov Chains to create fake Trump tweets

Comment

trumpitter

A Trump Tweet Bot that uses Markov Chains to create fake Trump tweets.

to-do

read in trump's new tweets
allow users to vote on the bot's creations
add highest upvoted bot generated tweets to the markov model
add a statistics page for most liked/favorited/retweeted
add Linux documentation

local setup

In order to set up trumpitter on your local machine, you will need ruby, and postgresql to be installed on your local machine.

windows

First install ruby using the ruby installer - https://rubyinstaller.org/
Next install the rails gem using gem install rails in your ruby terminal
Then install postgresql from https://www.postgresql.org/download/
You will need a tool to setup your local postgres servers so install pgadmin https://www.pgadmin.org/download/pgadmin-4-windows/
Launch pgAdmin and set up your username and password. Make sure you pass this in to ruby via the database.yml file so that you are able to use your local server.
You can now launch the server using ruby bin/rails server and can navigate to http://localhost:3000 to view the application.

macOS

First install ruby using homebrew with brew install ruby
Next install postgresql with brew install postgres
Install the rails gem with gem install rails
Launch the postgres server with brew services start postgres
Finally you can launch the local server with rails server and navigate to http://localhost:3000 to view the application.

linux/unix

TODO

contributors

Ishan Khatri - ishan@khatri.io

Jasmine Lee - leejas2017@gmail.com

Will Morgus - wmorgus@seas.upenn.edu

Nigel Paine - nigelspaine@gmail.com

Built With

Submitted to

PennApps XVI

Created by

I parsed Trump's tweets into a series of useable strings with a Python function, which I passed on to Jasmine. Next, I set up the Ruby on Rails project on which we deployed the service and added functionality, then spent way too much time making the front end just a little bit cheeky.

wmorgus Morgus
Worked on the weighted probability functionality for phrases so that all of Trump's tweets are analyzed and depending on the frequency of words used a certain probability would be generated to be able to choose the best words. Out of those weighted probabilities the algorithm would then come up with an output word. I also implemented a word frequency function for future stats and metrics.

Nigel Paine
Initially I planned out our project and came up with the algorithm by which we abstracted the Markov Chain concept into code. Then I wrote the core of the rails application to run the Markov model scripts when the user chooses to generate a new tweet.

Ishan Khatri
I wrote the hash table that takes in any .txt file and stores the dictionary of possible words. The dictionary for the real Trump tweets file is saved in a JSON file to save time. I then wrote the markov chains created based on the probabilities of word appearances stored in the dictionary to generate a tweet.

Jasmine Lee

Updates

Nigel Paine started this project — Sep 10, 2017 05:59 AM EDT

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.