{"id":"W2802736684","doi":"10.2196/publichealth.9361","title":"Clinical Relation Extraction Toward Drug Safety Surveillance Using Electronic Health Record Narratives: Classical Learning Versus Deep Learning","year":2018,"lang":"en","type":"article","venue":"JMIR Public Health and Surveillance","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":76,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"National Institute on Drug Abuse; National Heart, Lung, and Blood Institute; National Institutes of Health","keywords":"Artificial intelligence; Machine learning; F1 score; Computer science; Context (archaeology); Recurrent neural network; Artificial neural network; Support vector machine; Relationship extraction; Deep learning; Information extraction; Identification (biology); Natural language processing","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":["metaepi_narrow","sts","research_integrity"],"consensus_categories":[],"category_scores_codex":[0.01182103,0.0004332333,0.0009165546,0.0002868563,0.002060846,0.0003613289,0.0005943734,0.0002824292,0.00003577142],"category_scores_gemma":[0.002494759,0.0004449621,0.0001421947,0.001256969,0.0002694745,0.0009695477,0.0002651055,0.002926652,0.00003969119],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001087333,"about_ca_system_score_gemma":0.00273411,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0005738858,"about_ca_topic_score_gemma":0.002057998,"domain_scores_codex":[0.986075,0.008118291,0.001660076,0.001473974,0.0007178976,0.001954732],"domain_scores_gemma":[0.9944416,0.001774833,0.001386706,0.0006420047,0.0004755177,0.001279325],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"not_applicable","study_design_scores_codex":[0.0002706191,0.00007138339,0.5468946,0.0001764852,0.00002642318,0.000002745811,0.006659421,0.0001674005,0.000002146732,0.003661859,0.0003085831,0.4417584],"study_design_scores_gemma":[0.001537964,0.002436099,0.3090543,0.00005006197,3.382388e-7,0.00004287827,0.001008475,0.2964433,1.114831e-7,0.0001659826,0.3887585,0.000502057],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.410733,0.003376496,0.4997395,0.07886501,0.003432149,0.001362555,0.000003874068,0.00132943,0.001157993],"genre_scores_gemma":[0.9826776,0.00181731,0.01269215,0.001368202,0.001050134,0.00003522525,0.0000615,0.00005406482,0.0002438407],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.5719445,"threshold_uncertainty_score":0.9998002,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05990555228628694,"score_gpt":0.3916093414532457,"score_spread":0.3317037891669588,"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."}}