Generates new things Rush Limbaugh and Michael Bloomberg could say using a trigram Hidden Markov Model (with Dan Scarafoni)
Switch branches/tags
Nothing to show
Clone or download
Pull request Compare This branch is 10 commits ahead of scarafoni:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
personalities
planning
src
web
.gitignore
Procfile
README.md
con_text.txt
debate_script
lib_text.txt
punditry_app.py
punditry_main.py
requirements.txt
script.txt
send_messages.sh

README.md

Opinion_Bot

a n-gram markov chatbot that can keep track of it's internal probability transition table

Manifest

/planning- contains files illustrating the outline of the program and future design ideas/issues

Requirements

  • nltk
  • pymongo
  • bs4 (BeautifulSoup)
  • MongoHQ instance

Plan

  • Use MongoDB to store Markov chain data/political speeches (done)
  • Each time we analyze a new political speech, update the transition probabilities in Mongo and add a new "Speech" document (done)
  • Put each politician/radio host bot on a spectrum based on their opinions
  • Uses (potential final products):
    • Rush Limbot (generate new speeches in the style of a politician/radio host)
    • Have the bots complete a sentence (e.g. Rush Limbot: "I think Obama is ___")
    • Edit existing speeches in the style of a certain politician (e.g. make "Mitt Romney" version of an Obama speech)
    • Arguer that retorts things you say (or agrees with you) based on their opinions
    • Have each bot generate new words (combinations of other words) that have negative connotations of their opponents, explaining what their new word means (e.g. "sounds like ____")
    • Rate opinions/sayings of bots (like or dislike)