Precomputed per-word predictors for computational models of human sentence processing: word surprisal, several kinds of entropy, and unigram surprisal, scored over reading corpora (eye tracking, self-paced reading, N400, maze). Every row says what text unit it describes and what unit its number is in:

dataset tag predictor stimulus_id unit_index text_unit value value_unit

provo gpt2_ft_entropy_provo token_entropy provo_1 4 Apple 8.4213 bits… See the full description on the dataset page: https://huggingface.co/datasets/samuki-hf/psycholing-predictors.