Validation of geometric conceptual-space representations against expert knowledge

Develop quantitative evaluation protocols to validate geometric conceptual-space representations and trajectory-based concept recognition against human expert knowledge, establishing reliability and generalisability across large annotated datasets.

Background

The paper’s validation is qualitative, based on alignment between trajectory movement patterns and expert commentary in a small number of annotated games. The authors note the need for systematic evaluation across larger corpora.

They explicitly identify validation of geometric representations against human expert knowledge as an open research question, underscoring the importance of rigorous, scalable evaluation methodologies for conceptual spaces applied to abstract, temporally realised strategies.