{"id":"W4399919623","doi":"10.2196/57670","title":"Novel Approach to Personalized Physician Recommendations Using Semantic Features and Response Metrics: Model Evaluation Study","year":2024,"lang":"en","type":"article","venue":"JMIR Human Factors","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Preprint; Computer science; Information retrieval; World Wide Web; Data science","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":true,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001541035,0.0002044512,0.0002025305,0.0007972412,0.0004951774,0.0005803983,0.0004151265,0.00006139698,0.000005953267],"category_scores_gemma":[0.0001888113,0.0001849282,0.0000582266,0.001197683,0.00002420954,0.0004226205,0.0002353028,0.0003452026,0.000004477365],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002981354,"about_ca_system_score_gemma":0.0001595308,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001861802,"about_ca_topic_score_gemma":0.00002002186,"domain_scores_codex":[0.9976702,0.0004872697,0.0002395522,0.0006716724,0.0006604051,0.0002709121],"domain_scores_gemma":[0.9989375,0.0002668706,0.00006810038,0.0004463377,0.0001472987,0.0001338691],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0002035211,0.003109693,0.02332564,0.0008641639,0.0005224816,0.00001492641,0.7111635,0.09374558,0.0349662,0.05264593,0.005062994,0.07437538],"study_design_scores_gemma":[0.0002859953,0.0001538717,0.04760502,0.00005146294,0.00003196872,0.000004674415,0.001708006,0.9493949,0.00002564054,0.0003597364,0.0001505012,0.0002281719],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.860051,0.000112039,0.1374396,0.0006049002,0.0001828548,0.001186152,0.00001027508,0.0002393258,0.0001738697],"genre_scores_gemma":[0.9756607,5.119979e-7,0.02374169,0.0001718576,0.00005213772,0.0000953246,0.00001625071,0.00002543802,0.0002360879],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.8556494,"threshold_uncertainty_score":0.7541152,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1469689032351031,"score_gpt":0.4271232552693717,"score_spread":0.2801543520342686,"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."}}