Dissertation Defense Announcements

This dissertation explores the potential of natural language models, including large language models, to extract causal relations from medical texts, specifically from Clinical Practice Guidelines. The outcomes of causality extraction from Clinical Practice Guidelines on gestational diabetes are presented, marking a first in the field. We also release, the first of its kind, an annotated corpus of causal statements in the Clinical Practice Guidelines.
We address the challenge of classifying causal sentences with a small amount of annotated data at the inter-sentence level by treating it as a cross-domain transfer learning problem. Obtaining these classified sentences is the first step in extracting causality. Furthermore, we delve into the importance of modal verbs and the degree of influence from cause to effect. We show the capability of three models (BERT, DistilBERT, and BioBERT) to identify the degree of influence in the text.
Lastly, we tackle the challenge of sparse annotated data for the causality extraction from Clinical Practice Guidelines by, again, using transfer learning. We investigate the correlation between data similarity and the efficacy of transfer learning. We also investigate a zero-shot and few-shot approach to cross-domain transfer learning and quantify the link between data similarity and success rates. With the cross-domain few-shot transfer learning, we achieve an F1-score of 81%, which suggests transfer learning as a possible solution to address the limited availability of annotated data.

Peak load forecasting is crucial for reliable and effective grid operation. The day-to-day operation of the power grid requires load scheduling and dispatches of different energy resources including Energy Storage System and Demand Side Management programs. An effective implementation of these peak-shaving strategies relies heavily on when the peak demand occurs. Hence, forecasting the timing of peak load is as important as forecasting its magnitude.
A review of relevant literature indicates that there is no inclusive study on the topic of peak timing forecasting. This research aims to bridge the gap between industry requirements and academic research by addressing some key questions. First, the study defines the different forms of peak timing problems that arise in grid operation. Next, we investigate the problem of how we measure the peak timing forecast errors. The research critically reviews error measures used in the literature for peak timing forecasting. Based on the findings five new application-specific error measures are proposed. The research then focuses on one of the manifestations of the peak timing problem, that is, forecasting daily peak hours.
We analyzed the accuracy of peak hour forecasts from a state-of-the-art hourly load forecasting model and set it as the benchmark. The model selection process using different peak timing errors and load shape errors is investigated. Furthermore, two different frameworks for peak hour forecasting have been developed. The effectiveness of the proposed frameworks is empirically demonstrated in two case studies. The first case study is from a medium-sized Utility in the U.S. and the second one is from ISO New England. The proposed models demonstrate improved forecast results on the benchmark model by 12-16% in the test years of the two case studies. Additionally, when the models are only evaluated on the critical days with very high demands, they outperform the benchmark by 25-53%. Findings from this study emphasize the importance of developing explicit models for peak hour forecasting by analyzing the key determinants that vary with geographical location and regional factors.

There is a growing mental health concern among Black Ratchet queer womxn in educational and criminal justice realms and the Covid-19 pandemic has left the educational climate in a state of high stress and anxiety. Consequently, Black Womxn in education are quitting from burnout. According to research, teacher burnout has been a concern for more
than 30 years, but currently there is a crisis. According to research “This is a five-alarm crisis. We are facing an exodus as more than half of our nation’s teachers and other school staff are now indicating they will be leaving education sooner than planned.: (Jotkoff, 2022).
Covid-19 has caused an increase in depression, anxiety, suicide ideation, and other health concerns in this population. External hindrances including racism, sexism, homophobia, and other ideologies rooted in America; negatively influence their mental well-being. Black queer womxn and girls are restricted from obtaining proper access to mental health services that take into consideration how identities are critical factors in mental well-being. This analysis will provide a rationale for utilizing culturally relevant mindfulness practices for Black queer womxn in educational sectors.

Seasonal cold weather and frost action has a major effect on the design, construction, performance, and maintenance of roadways. Frost heaving and thaw weakening are especially problematic, subjecting all elements of a pavement system to significant changes in moisture content, stress, and strain. Nationally, this leads to recurrent annual maintenance costs estimated at over 2 billion dollars, as well as additional economic impacts because of related vehicle damage, road closures, and weight restrictions. Studies identify three basic requirements for frost action; freezing temperatures, availability of water, and frost-susceptible soils. While advances have been made in designing for freezing temperatures and providing for groundwater separation, very little progress has been made in terms of in situ soil improvement. A cost and labor-intensive approach is to replace unsuitable soils. As an alternative, Engineered Water Repellency (EWR), a process in which soils are made hydrophobic is presented as a suitable barrier limiting the transport of water through these soils, resulting in frost heave mitigation. This is achieved by combining soils with cost-effective and environmentally compatible polymers and other complex organic molecules. This process explored through a multi-year project funded by the U.S. National Science foundation, involved laboratory, field, and numerical studies.