Stimuli, scripts, and code for 'Structural Priming Demonstrates Abstract Grammatical Representations in Multilingual Language Models'
### Description of Content
* `cleaned_dfs` contains the experimental stimuli from the original studies, labelled with the relevant study, conditions, and languages. It also contains the relative production probabilities from the original human studies (in `human_results.tsv`).
* `cleaned_stims` contains the stimuli themselves, as used to run our analyses.
* `code` contains the Python code used to run our analyses (our specific analyses can be run using the `run_models.sh` script).
* `surprisals` contains the output from our language model experiments.
* `statistical_analysis` contains the statistical analysis scripts and figures.
* `test_contamination` contains the files necessary to test for language contamination in CC-100 as described in Appendix A of the paper.
