{"id":"W3035321115","doi":"10.18653/v1/2020.bionlp-1.19","title":"Extensive Error Analysis and a Learning-Based Evaluation of Medical Entity Recognition Systems to Approximate User Experience","year":2020,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":17,"is_retracted":false,"has_abstract":true,"ca_institutions":"National Research Council Canada","funders":"","keywords":"Computer science; Classifier (UML); Artificial intelligence; Machine learning; Judgement; Metric (unit); Test set; Medical diagnosis; Natural language processing; Gold standard (test); Information retrieval; Pattern recognition (psychology); Data mining; Mathematics","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":[0.01354224,0.001820553,0.001098386,0.002235485,0.0006373756,0.001442524,0.001163069,0.001712891,0.001618157],"category_scores_gemma":[0.04963029,0.0002754637,0.001000371,0.001540935,0.0008618836,0.002333777,0.001868255,0.001188813,0.0009252636],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001207667,"about_ca_system_score_gemma":0.0004892896,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003874345,"about_ca_topic_score_gemma":0.004733937,"domain_scores_codex":[0.984268,0.006696784,0.002266511,0.00258371,0.003641378,0.0005436747],"domain_scores_gemma":[0.9476908,0.03652547,0.002662371,0.005022915,0.007232179,0.0008661306],"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":[0.005951827,0.001949467,0.0971714,0.003019558,0.001963311,0.001261669,0.003162786,0.2094024,0.04857032,0.00212871,0.01974757,0.605671],"study_design_scores_gemma":[0.0002060515,0.005673803,0.1088815,0.0002838125,0.0004416679,0.001911833,0.001257722,0.7626787,0.1051831,0.002451265,0.01067538,0.0003551454],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.866016,0.002583394,0.1189614,0.0003551575,0.0003736332,0.0004284694,0.002866106,0.004723292,0.003692656],"genre_scores_gemma":[0.9433745,0.0003037546,0.04765461,0.0001374842,0.00006087236,0.0002150055,0.005478891,0.0002991192,0.002475814],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01354224,"threshold_uncertainty_score":0.07161903,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08982900735465392,"score_gpt":0.3426717351197669,"score_spread":0.252842727765113,"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."}}