By Kaylie Curran

Faculty Mentor: Dr. Debra Hydorn

Abstract

This project explores how the BERT (Bidirectional Encoder Representations from Transformers) language model embeds text based on word context. By analyzing excerpts from famous novels, the study examines how BERT generates different embeddings for the words. Techniques such as co-occurrence graphs, dot product similarity, and pairwise distance measurements are used to analyze the relationships between word embeddings. The goal is to gain insight into how BERT interprets words and how this influences AI’s ability to generate responses.