{"id":"W4385572948","doi":"10.18653/v1/2022.blackboxnlp-1.8","title":"Post-hoc analysis of Arabic transformer models","year":2022,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University","funders":"","keywords":"Computer science; Transformer; Natural language processing; Arabic; Artificial intelligence; Vocabulary; Semitic languages; Arabic languages; Linguistics; Speech recognition; Engineering","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":[0.002348162,0.001678643,0.0007401261,0.0008717016,0.000607739,0.001613507,0.001591162,0.0009447419,0.013386],"category_scores_gemma":[0.01385976,0.0004504937,0.001210254,0.0005772355,0.0005141988,0.001717276,0.001308097,0.003342508,0.004851875],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001546359,"about_ca_system_score_gemma":0.001846811,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.009942003,"about_ca_topic_score_gemma":0.01710713,"domain_scores_codex":[0.9989221,0.0002879103,0.00007164918,0.0002838015,0.0002150682,0.0002194597],"domain_scores_gemma":[0.9920902,0.004522664,0.0002015562,0.001107963,0.00187439,0.0002032223],"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":[0.001926599,0.0004393865,0.01966864,0.0005683283,0.00078552,0.000935643,0.000580248,0.3570418,0.03395705,0.01077726,0.04137215,0.5319474],"study_design_scores_gemma":[0.00004037334,0.0002136956,0.002914543,0.000034912,0.0001009895,0.0001181492,0.0002611804,0.9628738,0.02200809,0.006171106,0.005237862,0.00002538745],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5987428,0.002310383,0.3428039,0.00230136,0.001133624,0.0005499774,0.006603884,0.01945038,0.02610364],"genre_scores_gemma":[0.9295847,0.0002970044,0.04749692,0.0003943287,0.0001111537,0.0002434779,0.009032194,0.001368401,0.01147181],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.013386,"threshold_uncertainty_score":0.04478067,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02316346400162472,"score_gpt":0.2343240879897945,"score_spread":0.2111606239881697,"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."}}