{"id":"W4376113940","doi":"10.1259/bjr.20220769","title":"Transformer versus traditional natural language processing: how much data is enough for automated radiology report classification?","year":2023,"lang":"en","type":"article","venue":"British Journal of Radiology","topic":"Artificial Intelligence in Healthcare and Education","field":"Medicine","cited_by":30,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"National Institute of Biomedical Imaging and Bioengineering","keywords":"Artificial intelligence; Machine learning; Random forest; Computer science; Deep learning; McNemar's test; Medicine; Natural language processing","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0009874682,0.0001165124,0.0004161364,0.0001994606,0.0001852268,0.00003779054,0.0002667147,0.0002188179,0.00007333909],"category_scores_gemma":[0.001330769,0.0001194535,0.0001205376,0.0003090042,0.0001861892,0.0002945637,0.000008979121,0.0004028994,0.00001044309],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001157479,"about_ca_system_score_gemma":0.0008301081,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00004998942,"about_ca_topic_score_gemma":0.00003722561,"domain_scores_codex":[0.9983392,0.00009247669,0.000709958,0.0003025556,0.0002051155,0.0003507153],"domain_scores_gemma":[0.9982858,0.0004273073,0.0003927007,0.0002368342,0.0005080389,0.0001492661],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"case_report","study_design_scores_codex":[0.001160234,0.0001892267,0.001553834,0.0002140186,0.0002597817,0.001976592,0.002276844,0.000006792342,0.005203394,0.0001487424,0.7116002,0.2754103],"study_design_scores_gemma":[0.003542059,0.004179458,0.1000383,0.0006429436,0.0007388738,0.7326636,0.01775324,0.0720489,0.003644065,0.001622493,0.0623983,0.0007276715],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9389032,0.006919417,0.001371661,0.0476402,0.004045899,0.000548924,0.0002088308,0.0001866822,0.0001751996],"genre_scores_gemma":[0.9922665,0.0008906941,0.002013339,0.0005080383,0.001966631,0.00002482153,0.001872514,0.00002399206,0.0004334911],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.7306871,"threshold_uncertainty_score":0.4871174,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.3207568730192971,"score_gpt":0.4595928361469281,"score_spread":0.1388359631276311,"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."}}