{"id":"W4400898279","doi":"10.1111/jedm.12406","title":"Using Automated Procedures to Score Educational Essays Written in Three Languages","year":2024,"lang":"en","type":"article","venue":"Journal of Educational Measurement","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":7,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"","keywords":"Mathematics education; Natural language processing; Computer science; Psychology; Linguistics; Artificial intelligence; Philosophy","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.005157715,0.001032616,0.0006608579,0.003692593,0.0004969728,0.002299851,0.001045723,0.0006058632,0.004119306],"category_scores_gemma":[0.03298691,0.0002484437,0.0003722198,0.001465018,0.0003699122,0.00129249,0.001557643,0.0006799394,0.002691472],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006144473,"about_ca_system_score_gemma":0.001010972,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002250241,"about_ca_topic_score_gemma":0.004657552,"domain_scores_codex":[0.9920464,0.003361837,0.0008698298,0.001234125,0.002262732,0.0002251882],"domain_scores_gemma":[0.9688612,0.01304521,0.003167753,0.002100682,0.01195374,0.0008714346],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0008439936,0.0005127,0.04003774,0.000469165,0.0001557472,0.0003305466,0.00131051,0.006618808,0.04646496,0.0009735536,0.004463994,0.8978182],"study_design_scores_gemma":[0.0005791047,0.00287867,0.2662421,0.0004202122,0.0003353503,0.002128891,0.005289431,0.4717295,0.2102603,0.005865796,0.0337549,0.0005157078],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7465224,0.0003807118,0.2235245,0.0002522572,0.0002095983,0.001177902,0.00218063,0.01197834,0.01377381],"genre_scores_gemma":[0.818852,0.0001523614,0.1724462,0.00006588506,0.00007373051,0.0004400876,0.002106361,0.0002581722,0.005605296],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.005157715,"threshold_uncertainty_score":0.02727687,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0603400789110201,"score_gpt":0.3634531269538653,"score_spread":0.3031130480428452,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}