Mapping assessment practices in DanceSport: a Human-In-The-Loop systematic review.

BACKGROUND: DanceSport is an aesthetic sport that combines athletic execution, artistic expression, musicality, and partner interaction. Although formal adjudication remains central to DanceSport competition, recent studies have increasingly explored technology-assisted approaches to provide more measurable and traceable evidence. However, the field lacks a focused synthesis explaining what major themes have emerged, how these themes have developed, what technologies have been applied, and how existing systems have been validated.

METHODS: This study conducted a systematic review enhanced by Human-In-The-Loop topic modeling. Following PRISMA-informed procedures, 1,957 records were retrieved from the Web of Science Core Collection, 192 full-text articles were assessed for eligibility, and 24 studies were included in the final review. Exploratory LDA topic modeling was applied to study abstracts to identify major themes and similarity-based thematic transitions, while qualitative content analysis of full texts examined assessment approaches, technologies, dance-category coverage, and validation designs.

RESULTS: Four major themes were identified: 3D dance motion perception, dance posture and motion evaluation, image analysis of dynamic dance performance, and validation of dance evaluation. Together, these themes reveal a research landscape that is increasingly centered on objective and measurable performance evidence. Building on earlier concerns with judging validity, the field has gradually shifted toward technology-assisted and data-driven assessment, with research focusing primarily on indicators such as posture, motion, balance, rhythm, and movement recognition. To support this transition, studies have increasingly employed motion capture, wearable sensors, computer vision, and computational models. Nevertheless, despite these technological advances, validation remains limited by uneven dance-category coverage and a continued reliance on isolated movements, short performance clips, or laboratory-based tasks.

CONCLUSIONS: DanceSport assessment research is moving toward technology-assisted and multimodal evidence systems, but technical measurement has not yet been fully integrated with aesthetic judgment, expert scoring, and competition-level validation. Future research should develop cross-dance datasets, compare Standard and Latin dances, test complete competition routines, examine criterion validity between technical indicators and expert scores, and involve judges, coaches, and athletes in assessment frameworks.