Author: Zion Thomas

One of the main problems that the Humor Genome Project currently has is interacting large scale with the Georgia Tech community. I believe that with Tech’s recent focus on mental health and suicide prevention the Humor Genome Project may be able to interact with the school administration to build some sort of comic relief for students. It has been proven in multiple studies that humor positively affects humans mental health and general well being. Because of this, I think that the project could serve as an avenue for students to get relief from studying and finals. Although the project is not large scale now, once the team is up and running it could be possible to collaborate with students who are interested in comedy, as well as actual comedians and releasing the finds of the team’s studies. Whether it be through events hosted on campus, or open sessions of the class that other students could come to.

This semester, the Crawler Team was building upon the website structure created in past semesters for the humor genome project. The main goals for this semester was to get the application hosted live online while still working as it did previously locally. The team was able to effectively get the site online hosted with heroku at https://humor-genome-project.herokuapp.com. In addition to this, the site is able to accept new users for login/sign up functionality. The webapp is linked with the MongoDB database containing all of the jokes that have been collected. As a privileged user, you are able to rate jokes and edit them and their content in the database. The site now has a jokeID for every joke, which makes it easier to reference the jokes in the database and also has a more aesthetic layout. For the Georgia Tech hosted PWP site, the team updated all of the member pictures and fixed small bugs with the connection to Twitter. For the future, the team plans to provide documentation for the heroku site, expand the web app to get a larger data sample size, use the site for analytics, and also get more activity on the social media and blog sites.

The Analytics Team did a lot of good work this semester for the project. They used various algorithms to collect data on jokes and attempted to create their own using machine learning computation. They sent out multiple class surveys to the entire class in an effort to recommend similar jokes to users. To do this, the team used Item-Item collaborative filtering and User-User collaborative filtering. One of these looked at the similarities between different jokes and the other looked at the similarities between different users’ interests.

The analytics team used a neural network on a collection of Reddit jokes, referenced at http://karpathy.github.io/2015/05/21/rnn-effectiveness/. It trained for about forty hours and the results that it gave were humorous. Here’s the joke:

“My wife picks up his own finger. One of them is spending the church. “Go immediately pointing,” and every morning “I have a honeymoon. “How can I hear?”.”

As you can see this joke makes very little sense, but it was interesting to see how a neural net algorithm could work to generate new material. In upcoming semester, the team plans to tune the algorithm with different identifiers to yield better results.