We Used the word ‘apeshit’ in a Paper for a Very Important Conference 🙈

The International Conference for Machine Learning (now in its 36th year — hard to believe ICML has been around for that long!), takes place June 5–9 in Long Beach, California this year.

As you might have gleaned from our previous post about ICLR, getting a paper accepted into a major ML conference is a Big Deal, in fact, it is the primary way scientists share their research with others in their field. 😇

Limor Gultchin, who will be presenting this paper at ICML

This year, our CSO Geneviève Patterson had a paper accepted! She co-authored it while she was at MSR) with Limor Gultchin (Oxford), Nancy Baym, Adam Tauman Kalai (MSR), Nathaniel Swinger (Lexington HS).

If you’re at ICML, Limor will be giving a talk on Thursday, June 13 at 12:05 pm, room 104. If you watch, tweet @thetrashapp!

Humor in Word Embeddings

Read the 10-page paper here: Humor in Word Embeddings: Cockamamie Gobbledegook for Nincompoops. This paper explores the computational representation of humor in single words and bigrams. Abstract:

While humor is often thought to be beyond the reach of Natural Language Processing, we show that several aspects of single-word humor correlate with simple linear directions in Word Embeddings. In particular: (a) the word vectors capture multiple aspects discussed in humor theories from various disciplines; (b) each individual’s sense of humor can be represented by a vector, which can predict differences in people’s senses of humor on new, unrated, words; and c) upon clustering humor ratings of multiple demographic groups, different humor preferences emerge across the different groups. Humor ratings are taken from the work of Engelthaler and Hills (2017) as well as from an original crowdsourcing study of 120,000 words. Our dataset further includes annotations for the theoretically-motivated humor features we identify.

Through extensive human experiments and neural network embedding space manipulations, Geneviève and her co-authors explored 3 main questions:

1. Can we use word embeddings to capture humor theories and identify a ‘humor direction’?
2. Can we identify different senses of humor across demographics
3. Can we define individual senses of humor and predict users’ taste?

Here’s a taste of how the data stacked up against those questions:

lololol how could you even do research on this right? You would just be lmao nonstop with your colleagues. I mean, I’m dead just looking at these word groups. 😂 😭 💀 Get in, and have some laughs with your science! 💩

– Geneviève & Team TRASH