{"id":"W6921374466","doi":"10.6084/m9.figshare.c.4029727.v1","title":"Technology-assisted title and abstract screening for systematic reviews: a retrospective evaluation of the Abstrackr machine learning tool","year":2018,"lang":"en","type":"other","venue":"Figshare","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Workload; Interquartile range; Task (project management); Health records; Sample size determination; MEDLINE; Citation; Range (aeronautics)","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":false,"about_ca":true,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":["metaresearch"],"category_scores_codex":[0.3345482,0.001560127,0.003654035,0.02023412,0.001134051,0.004531408,0.003723456,0.001350711,0.004932071],"category_scores_gemma":[0.6514934,0.002432583,0.005380274,0.02000795,0.001549341,0.004968734,0.00548604,0.001572281,0.002069372],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.004267347,"about_ca_system_score_gemma":0.01846498,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00435437,"about_ca_topic_score_gemma":0.01079719,"domain_scores_codex":[0.6023902,0.21922,0.1166664,0.01270101,0.047641,0.001381355],"domain_scores_gemma":[0.1286862,0.6150308,0.1125937,0.05347351,0.0879701,0.002245678],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.00879276,0.0007102502,0.1225419,0.1580805,0.01306798,0.0006762341,0.00860229,0.004145987,0.002968144,0.001655937,0.02559092,0.6531671],"study_design_scores_gemma":[0.01457615,0.02160008,0.4545735,0.1632123,0.06515582,0.008165133,0.00633613,0.07705119,0.01661961,0.005279066,0.1643497,0.003081313],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.4942592,0.115962,0.1697408,0.00994266,0.001052675,0.1102884,0.07537073,0.01074297,0.01264054],"genre_scores_gemma":[0.5533993,0.01942587,0.3310727,0.002144733,0.0005084191,0.07172789,0.01911811,0.001149544,0.001453381],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6654518,"threshold_uncertainty_score":0.8206206,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1255207064563569,"score_gpt":0.3545363553996457,"score_spread":0.2290156489432887,"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."}}