Publications / Robyn Correll, MPH, CHWI

Preprint • 3rd June 2025

Practical Considerations for Fine-Tuning BERT-Based Language Models in Health Research: Lessons from Classifying Anti-Vaccine Posts on Social Media

For those transitioning from traditional machine learning to LLMs, it is crucial to understand how choices in data collection and training-including keyword selection, fine-tuning data quantity, stopping conditions, collapsing categories, and labeling conflicts-can significantly impact model performance, especially with smaller datasets.

Screenshot of the preprint, including title, authors, and funding information

Preprint • 14th July 2023

Classification Performance Thresholds for BERT-Based Models on COVID-19 Twitter Misinformation

To examine the effects of keyword selection and training data size might impact model performance, two separate Twitter datasets relating to sentiments regarding COVID-19 vaccination — one using a more specific set of keywords and another using more generalized keywords — were used to train two BERT models with increasing quantities of hand-labeled data.

Multi-colored bar graph showing the breakdown of COVID vaccine misinformation by topic area. The longest bar is

BMJ Open • 28th November 2022

COVID- 19 vaccine misinformation in English- language news media: retrospective cohort study

To examine the magnitude of vaccine misinformation in traditional media (Professionally Generated Content, not User Generated Content), we used Cision’s NextGen Communications Cloud platform12 for the period 27 July 2020 (the day that phase 3 trials for the Moderna and Pfizer vaccines began) through 30 June 2021.

The Immunization Partnership • 14th January 2019

Texans Unite to Protect: Working Together to Eliminate Vaccine-Preventable Diseases

After analyzing data collected from hundreds of immunization stakeholders at town hall meetings and surveys, scouring peer-reviewed journals, and crunching data sets, this report outlines the most important challenges and opportunities in vaccine-preventing and emerging diseases threatening Texas today.

The Immunization Partnership • 6th December 2016

A Time for Action: Recommendations for Improving Texas Immunization Rates (2016)

This publication is the culmination of more than six months of research – gathering input and feedback from hundreds of immunization stakeholders at town hall meetings and via a survey on the challenges and opportunities they experience as they work to protect Texas communities from vaccine-preventable diseases. The feedback collected, as well as information gathered from state and national government agencies and peer-reviewed papers, is presented in this report as a means to shed light on the state of immunizations in Texas, as well as serve as a road map for all immunization stakeholders – healthcare providers, public health personnel, and legislators alike – on what can be done to protect and improve immunization rates in Texas communities.