Systematicity in GPT-3’s Interpretation of Novel English Noun Compounds
This is the repository containing all annotations and data created in the paper. The Experiment 1.xlsx
, Experiment 2.xlsx
, and Experiment 3.xlsx
correspond to (1) replication of Levin et al. (2019) experiments on GPT-3 Davinci, (2) extension of (1) using additional novel noun compounds, and (3) probing experiment
conducted by replacing compound components with nonsense strings.