As an intern in the AI in Medicine (AIM) lab at BWH, I helped to create and use an annotation schema for social determinants of health (SDoH) in clinical notes. I also deployed novel strategies using the language models GPT-3.5 and LlaMA to generate and classify data, and began investigations exploring the potential biases and harms of GPT-3.5.

In this role, I was part of a team building a corpus of verb phrase ellipsis (VPE). In my first semester, I annotated instances of ellipsis; in the second, I reviewed and discussed literature on VPE and on annotation of theoretical syntax phenomena with my colleagues and created a pilot annotation schema for the linguistic details of these instances of VPE.

In this role, I was part of a team annotating the arguments of verbs: agents, themes, instruments, results, and locations. I quickly learned annotation schema for associating verbs with their arguments, and I actively engaged with project directors to ensure accuracy and consistency of annotations.

In this role, I collaborated with peers and mentors to research the correlations between a person's prosody and their performance on automatically-graded job interview task. We evaluated sources of bias in interview scores, and the effects of this bias on SVM models. During this internship, I made extensive use of Praat, Parselmouth, and MatPlotLib. I attained a practical knowledge of experimental statistics and the research presentation process; along the way, I also earned a Dean's Citation for Broadening Participation in Research by reaching out to groups underrepresented in data science