• Axel Strubel
• Avi Barazani
• Yaniv Ben-Malka

Using machine learning tools, we have built a supervised nlp-based jokes regressor, that gets an English joke as an input, creates various features that try to capture the sense of the joke, and predicts how funny it is, in a scale 0 to 100.

A reddit jokes dataset, that contains more than 180,000 jokes and their score (votes).

Our future intention is to use the scrap tweets from a pre-made list of 'funny' pages, and use the tweets to further train & test the model, in addition to some more twitter-based feature engineering.

Notebooks

• EDA
• Feature Engineering
• Feature Processing (Dimensionality Reduction)
• Modeling

Joke_Predictor

• predict_joke.py (Run the entire pipline for predicting an input joke score)

Data

• reddit_jokes.json - Dataset currently in use

Twitter_API (Not in use)

• Get_Tweets.py - Tweets scraper
• tweets.csv - A dataset of scraped tweets

docs.google.com/document/d/1f_8DQBT13SHORA22K4gOfqOf_l_9iXPtuVgdrXE55Cw