{"id":"W3180779372","doi":"10.1038/s43856-021-00008-0","title":"A BERT model generates diagnostically relevant semantic embeddings from pathology synopses with active learning","year":2021,"lang":"en","type":"article","venue":"Communications Medicine","topic":"COVID-19 diagnosis using AI","field":"Medicine","cited_by":19,"is_retracted":false,"has_abstract":true,"ca_institutions":"Juravinski Hospital; University of Waterloo; McMaster University","funders":"","keywords":"Computer science; Artificial intelligence; Deep learning; Natural language processing; Information extraction; Context (archaeology); Semantics (computer science); Training set; Information retrieval","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":[0.0002789774,0.00026652,0.0006633924,0.0001639725,0.000358778,0.00002261487,0.0004325789,0.0001452657,0.0002375717],"category_scores_gemma":[0.006154717,0.0002128713,0.0000670425,0.0005147446,0.0007326568,0.00009396182,0.0004440995,0.0008435262,0.00005403223],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001859814,"about_ca_system_score_gemma":0.0005609155,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002972903,"about_ca_topic_score_gemma":0.0002699908,"domain_scores_codex":[0.9980462,0.0003022071,0.0004567503,0.000507692,0.0003581627,0.0003289635],"domain_scores_gemma":[0.9922112,0.004445232,0.0001806616,0.002209495,0.0007163485,0.0002370823],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"bench_or_experimental","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001432999,0.007308804,0.08145989,0.0008400896,0.002484704,0.004841567,0.05100489,0.01986198,0.6570895,0.01038723,0.07024148,0.09304684],"study_design_scores_gemma":[0.01793372,0.003408751,0.07475836,0.015676,0.00734476,0.002061027,0.01139079,0.6814549,0.0556801,0.005231902,0.1230121,0.002047614],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5840228,0.01074129,0.01864696,0.3838884,0.0001150534,0.0005354304,0.00002992314,0.0003745118,0.001645586],"genre_scores_gemma":[0.9319075,0.006880797,0.04130315,0.0180274,0.0001574969,0.000166163,0.000722344,0.0000735439,0.0007615917],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6615929,"threshold_uncertainty_score":0.8680639,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04592017625887133,"score_gpt":0.3451619568347017,"score_spread":0.2992417805758303,"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."}}