{"id":"W6948191620","doi":"10.48448/ztc5-5r72","title":"What's in the Box? An Analysis of Undesirable Content in the Common Crawl Corpus","year":2021,"lang":"en","type":"other","venue":"Underline Science Inc.","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Mila - Quebec Artificial Intelligence Institute; Université de Montréal","funders":"","keywords":"Language model; Corpus linguistics; Computational linguistics; Content (measure theory); Language identification; Training set; Artificial neural network","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.00224854,0.0002993197,0.0003927215,0.004192722,0.002059986,0.001888809,0.0005321075,0.0007105776,0.002984155],"category_scores_gemma":[0.01861675,0.0002787298,0.000255183,0.004595523,0.001779541,0.001581024,0.001734155,0.001056094,0.001689714],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001085509,"about_ca_system_score_gemma":0.001638987,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01475515,"about_ca_topic_score_gemma":0.04148374,"domain_scores_codex":[0.9977922,0.0008085566,0.0001900441,0.000324282,0.0007545832,0.000130227],"domain_scores_gemma":[0.981187,0.01107152,0.001113257,0.002723444,0.003512607,0.0003921792],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"observational","study_design_scores_codex":[0.0006962416,0.0005288121,0.1449001,0.002414638,0.0001437368,0.004753802,0.02757055,0.003718752,0.02691497,0.04606503,0.317748,0.4245454],"study_design_scores_gemma":[0.00008913383,0.0001875206,0.2642845,0.0009873949,0.0001700143,0.006243821,0.01859102,0.03436603,0.03738867,0.01545639,0.6220505,0.0001849434],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8665257,0.002307683,0.03976195,0.008028152,0.0004758281,0.0003381703,0.03507889,0.00251431,0.0449693],"genre_scores_gemma":[0.8528374,0.001655542,0.04677344,0.002139439,0.0003250947,0.0006114482,0.07181305,0.00275021,0.02109437],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.01475515,"threshold_uncertainty_score":0.02933854,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1021472914619384,"score_gpt":0.3475380771891824,"score_spread":0.245390785727244,"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."}}