This is the data from "JBLiMP: Japanese Benchmark of Linguistic Minimal Pairs" (Someya and Oseki, 2023). For details see the original git repo or the paper.