Large-Scale Bitext Corpora Provide New Evidence for Cognitive Representations of Spatial Terms

Explore groundbreaking research on cognitive representations of spatial terms in this 15-minute conference talk presented at EACL 2024. Delve into the first corpus-based empirical test of the hypothesis that geometric and functional spatial terms exhibit different cross-linguistic variability. Learn about the innovative pipeline developed to extract, isolate, and align spatial terms in basic locative constructions from parallel text. Discover how Shannon entropy is used to measure variability of spatial term use across eight languages, and examine the findings that support the significant difference between functional and geometric terms. Gain insights into the latent variable modeling that further reinforces the division of spatial terms into geometric and functional classes. Understand the implications of this research for cognitive science and language learning, particularly in relation to the mastery of spatial terms over time.