Fine-Tuned Word Embeddings for Enemy Item Detection

• In Event: Applications of AI

Abstract

This study evaluates GloVe 6B and Clinical BERT for detecting enemy items in a certification item bank. Using word embeddings and machine learning, we aim to improve test precision and validity. Fine-tuned embeddings on specialty corpora contain context relevant to the specialty area with the potential for improved prediction accuracy.

Authors

• Liu Liu, University of Washington

• Marcus Walker, National Commission on Certification of Physician Assistants