While brainstorming for this hackathon, a certain teacher we have came into mind, and we thought about how nice it would be to be able to replicate his spoken habits.

What it does

Impressionator seeks to take existing sets of text, such as from Donald Trump's Twitter account or The Bible, and generate short blurbs that could have been part of that text. Impressionator also supports combining two different entities (out of 5) and generating a hybrid of the two (Donald Trump's bible?)

How we built it

We used Node.js to create a full-stack app, standard HTML/CSS for the front-end, and Python to run the machine learning and generate these new texts.

Challenges we ran into

We found that many libraries that we wanted to use were not greatly (if at all) documented, and I feel like we all gained a further appreciation for well-documented libraries/APIs.

Accomplishments that we're proud of

Honestly, getting anything to work was extra satisfying, whether it was as big as the first generated text or as small as making flex-box work for the homepage.

What we learned

We learned a lot in various fields, from using Ajax for asynchronous calls, running Python in Node, and implementing Markov chains.

What's next for Impressionator

Implementation of Neural Networks and the implementation for a more general impression: users would be able to add their own text and corpuses to make impressions on. This way, they could make impressions for entities not currently implemented.

Share this project: