Weinberg, Eric

My current research project tackles the complexities of university course timetabling, a puzzle that has intrigued scholars and administrators alike. This project is a foray into the application of computational algorithms to create efficient, conflict-free academic schedules accommodating diverse course credits and faculty availabilities. This research focuses on developing a new approach to university course scheduling using a Reconstructive Multi-Objective Hybrid Mixed Integer Genetic Algorithm. This method combines heuristic strategies with Mixed Integer Programming (MIP) and Genetic Algorithms (GAs) to tackle the complex challenges of course timetabling. By reorganizing schedules around established academic time-slot patterns, the approach aims to reduce complexity and improve scheduling flexibility, providing a more adaptable solution for academic institutions. This work explores how these algorithms can offer practical improvements over traditional scheduling method

My research on COVID-19, showcased through the Balefire.info Covid social data explorer, stands at the confluence of epidemiology and data science. By meticulously tracking the longitudinal data of the pandemic alongside related social factors, this project offers a granular view of the virus's impact on communities. The platform, built with collaboration from various data providers, harnesses sophisticated data analytics to uncover trends and correlations that inform public health decisions. The insights drawn from this research not only contribute to our understanding of COVID-19's societal effects but also enhance our preparedness for future public health challenges.

Modeling Meaning in the Early Republic is a digital humanities project that applies modern natural language processing techniques—especially transformer-based large language models (LLMs)—to the study of political discourse in the Founding Era. By fine-tuning LLMs on 18th-century American texts, the project generates historically grounded word embeddings that capture how meanings of key terms like freedom, property, and democracy were shaped by their original context. This allows for the study of semantic drift over time and across authors, revealing subtle shifts in how early American thinkers framed concepts of power, labor, and citizenship. The project demonstrates a novel use of contextualized embeddings for diachronic analysis, bridging computational linguistics and historical inquiry. It contributes to the growing field of historically informed NLP by showing how transformer architectures can be repurposed not just to generate text, but to recover and visualize the conceptual structure of past political language.