{"id":"W2074156882","doi":"10.1109/slt.2012.6424163","title":"Realistic answer verification: An analysis of user errors in a sentence-repetition task","year":2012,"lang":"en","type":"article","venue":"","topic":"User Authentication and Security Systems","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Computer science; Sentence; Task (project management); Repetition (rhetorical device); Imperfect; Simple (philosophy); Authentication (law); Word error rate; Natural language processing; Speech recognition; Artificial intelligence; Human–computer interaction; Computer security","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":[0.0004430532,0.00006605915,0.000156517,0.0003569719,0.00002514178,0.00003425731,0.0003370545,0.00003965741,0.00007996517],"category_scores_gemma":[0.00002545698,0.00005962379,0.00005735504,0.001296332,0.00002630459,0.000651229,0.00003027834,0.00003886841,0.0000269779],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00003017442,"about_ca_system_score_gemma":0.0000172487,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000806322,"about_ca_topic_score_gemma":0.0005591785,"domain_scores_codex":[0.9990104,0.0001182635,0.000301054,0.0001967673,0.0002211883,0.0001523407],"domain_scores_gemma":[0.9990731,0.00002799972,0.00009790849,0.0006251134,0.00008740141,0.0000884304],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"observational","study_design_scores_codex":[0.000008254856,0.0008855886,0.3880342,0.0000348749,0.0001655163,0.000001597609,0.09748025,0.0001605972,0.002815113,0.5091422,0.000274244,0.0009975486],"study_design_scores_gemma":[0.0001326997,0.00002039044,0.6338966,0.000008984615,0.00006739442,0.00000199237,0.0007139692,0.3631009,0.0003375321,0.0002089835,0.001384647,0.0001258611],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.8071624,0.00002531427,0.1905407,0.0004140103,0.0002101412,0.0001335772,0.000006238265,0.0000722195,0.001435405],"genre_scores_gemma":[0.9979795,0.000003857985,0.001596595,0.0001149635,0.00001894815,0.00001179962,0.00003119859,0.000002702938,0.000240472],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.5089332,"threshold_uncertainty_score":0.2431388,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02646676874574957,"score_gpt":0.281795902712501,"score_spread":0.2553291339667514,"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."}}