{"id":"W4407626533","doi":"10.1016/j.mlwa.2025.100630","title":"Predicting classification errors using NLP-based machine learning algorithms and expert opinions","year":2025,"lang":"en","type":"article","venue":"Machine Learning with Applications","topic":"Imbalanced Data Classification Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University; Western University","funders":"","keywords":"Artificial intelligence; Computer science; Machine learning; Natural language processing","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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0004852275,0.0002894256,0.000256997,0.0004382518,0.00121225,0.0002803454,0.0007576906,0.0001224296,0.000007949267],"category_scores_gemma":[0.0001582668,0.0002745907,0.00004655947,0.001387984,0.0001625835,0.0004343016,0.0002481875,0.0007621329,0.00000873493],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001413775,"about_ca_system_score_gemma":0.0001825319,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002902428,"about_ca_topic_score_gemma":0.0000437133,"domain_scores_codex":[0.9978397,0.000213712,0.0004319111,0.0008512286,0.0003022584,0.0003611568],"domain_scores_gemma":[0.9981847,0.0002872014,0.000352913,0.0008640808,0.0001868836,0.0001241774],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00005618697,0.0005429441,0.4896992,0.000221333,0.0001629521,0.00000480424,0.001234299,0.02263068,0.02262332,0.1707624,0.0003583065,0.2917035],"study_design_scores_gemma":[0.0004242031,0.00006717618,0.01072106,0.00008437423,0.00002005645,0.00001230789,0.00007497886,0.931987,0.001347157,0.0003303356,0.05465633,0.0002749976],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.001473413,0.0004070488,0.9914439,0.003446928,0.00004617346,0.000725797,0.00002055761,0.001503423,0.0009327297],"genre_scores_gemma":[0.5824428,0.00005549599,0.4153998,0.0003248868,0.00005179716,0.0008125415,0.0004759435,0.00003817008,0.0003985632],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9093564,"threshold_uncertainty_score":0.9999706,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02403382708901836,"score_gpt":0.3024235520663833,"score_spread":0.278389724977365,"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."}}