Make Slack Great Again
A responsive chatbot which mimics political rhetoric and teaches American history
Inspiration: This election season has seen the most absurd political comments since since the creation of ENIAC. To demonstrate the rhetoric’s simplistic patterns - especially from Republican nominee Donald Trump - we set out to make a chatbot which could pass as modern politician. With chat support for Franklin D. Roosevelt, Bernie Sanders, Donald Trump, and Sonia Sotomayor we hoped to demonstrate the full gamut of American politics. What began as a tool to mimic political rhetoric quickly expanded when we realized the uses that our technology has in a classroom.
Our Bots are perfect for teaching tech savvy students about the policies and personalities which govern the country. Sometimes funny, sometimes inspirational, and always a blast to talk to - we are having a hard time picking a candidate! We are able to quickly add new personalities and expand the knowledge of existing characters. With our fine tuned machine learning chatbot, this system is easily scaleable.
The project is primarily a SlackBot framework, with support for voice through Amazon Echo. Users choose a bot in Slack and can ask questions in plain English which are quickly answered from a library of relevant quotations or from newly generated phrases in the style of the character. Trump and Sanders are trained for political discourse and can engage in moderated debate with each other while Roosevelt and Sotomayor are examples of the educational uses for the software and are focused on The New Deal and The US Constitution respectively.
Educational application: The Roosevelt and Sotomayor characters may have less mass appeal, but to us they are the more exciting test cases. The best application for our chatbots are in the elementary school classroom as a gateway to new worlds. Rather than read a history textbook, or memorize the names or supreme court justices, students could engage with the interactive personalities and ask questions for themselves. A few chats with Sotomayor will inspire any student in ways that textbooks can not. On top of that, Sotomayor acts as a shell for our legal encyclopedia. She answers questions about the law and about her childhood. This combination brings an entire interactive legal experience.
What we learned: This project crystallized for us the difference in how our politicians speak. From FDR’s motivational quotes, to Trump’s inflammatory criticisms, this taught us a great deal about modern political science. On the technical side we implemented Latent Semantic Indexing (LSI) to read and “understand” the questions that we provide. The concepts generated in that process are then matched to our human profile characteristics.When we are unable to find an appropriate response, our algorithm uses Markov chaining along with a probability equation and a check on grammar in order to weed out the worst answers and the ungrammatical ones. The sum of our Natural Language Processing (NLP) algorithms gives us a powerful and reactive personality.