A sample of chat with Deborah

Deborah Downer: Retro First World Problems

Kathy Lau, James Xu, Janet Zhang, Jason Mohabir

PennApps Spring 2018

Inspiration

We are inspired by famous chatbots and AIs such as GLaDOS and TayTweets. We wanted to write our own bot that would be able to give sarcastic, retro world problem responses to show how far technology had changed since to have some fun.

What it does

Deborah Downer is a Retro chatbot capable of understanding basic English phrases and communicating with the user. She provides some snarky commentary about the decades of past in response to first world problems. Deborah can be triggered using certain keywords to use media to help understand her responses better. The more Deborah is used, the smarter she becomes.

How we built it

Deborah is built in Python using the ChatterBot library. After using the ChatterBot machine learning library, Deborah was taught and trained using various examples of pop culture in the prior decades and provided conversations about various topics through the Ubuntu Corpus.

Challenges we ran into

Deborah is limited to only using responses that have been given to her in the past. She can not formulate new thoughts, only match previous ideas to current ideas using comparison algorithms. On top of this, Deborah has a tendency to not be comprehensible if its database is not wiped after tinkering with the Flask apparatus.

Accomplishments we are proud of

Deborah is knows snippets of information about various fields and can answer most clear questions with answers that it is at least 65% sure have relevance. On top of this confidence, Deborah is able to display media to convey its messages better.

What we learned

The team learned how to work with Python Libraries and how to work with Bots. The documentation on ChatterBot provided a cursory look into Machine Learning and Natural Language Processing.