Date of Award

Spring 6-3-2026

Document Type

Thesis (Undergraduate)

Department

Computer Science

First Advisor

Professor Nikhil Singh

Abstract

Powwow dance is a form of Indigenous expression that combines movement, storytelling, regalia, and community values. Within competition powwows, dancers are evaluated through a highly subjective judging process that often lacks standardized criteria, creating challenges for consistency, transparency, and fairness. This work investigates whether a learning-to-rank framework can model subjective powwow dance preferences while remaining grounded in the cultural context of the practice. To support this research, a new dataset was constructed consisting of 19 Women's Fancy Shawl dancer profiles and 236 pairwise preference labels collected from experienced members of the powwow community. Each dancer profile combines textual descriptions, images, and video, which are fused into a unified representation. These representations are then used within a RankNet-inspired pairwise learning-to-rank framework to predict human preferences and generate dancer rankings. The proposed system is evaluated using both held-out rater and pair-group validation strategies, achieving pairwise accuracies of up to $69.4$\% and Kendall Tau correlations exceeding $0.40$. In addition to predictive performance, this work introduces consistency analyses examining agreement across raters, ranking variance, and top-k exposure. The results demonstrate that learning-to-rank systems can capture meaningful patterns in subjective human evaluations and serve as a proof of concept for applying machine learning to culturally grounded and highly subjective ranking tasks. While not intended to replace human judges, the proposed framework highlights the potential of machine learning ranking systems to support more transparent and consistent ranking analysis.

Download

Share

COinS