In 2024, I was invited to present a guest lecture in the Human Language Technologies course at the University of Pittsburgh. Students in this course had recently learned about static word embeddings, but amidst the hubbub of news about generative AI, it felt necessary to provide additional context for why these “old-fashioned” models might be useful.
To demonstrate this, I presented Gennaro and Ash’s excellent 2022 manuscript exploring rhetoric in US congressional debates. This paper showed students an important application of static word embeddings outside mainstream NLP.
I presented an updated version of this lecture at the NLP course at Pitt in 2025. The slides for the 2025 version are included below. Thank you Lorraine Li for this generous invitation!