{"id":"W4226316739","doi":"10.2196/34834","title":"Pretrained Transformer Language Models Versus Pretrained Word Embeddings for the Detection of Accurate Health Information on Arabic Social Media: Comparative Study","year":2022,"lang":"en","type":"article","venue":"JMIR Formative Research","topic":"Topic Modeling","field":"Computer Science","cited_by":7,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Taibah University; Science Foundation Ireland","keywords":"Computer science; Social media; Natural language processing; Leverage (statistics); Language model; Artificial intelligence; Transformer; Arabic; Health informatics; Machine learning; Information retrieval; World Wide Web; Linguistics; Medicine; Public health","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":true,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["sts"],"consensus_categories":[],"category_scores_codex":[0.003642813,0.0001557788,0.0002815994,0.0004108594,0.001361651,0.0001241583,0.000921489,0.00004386809,0.00001440259],"category_scores_gemma":[0.00009373495,0.0001190823,0.0001043739,0.001064821,0.00009523256,0.00180253,0.0001840832,0.0007432583,0.00000446645],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004810781,"about_ca_system_score_gemma":0.0002718591,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00005614828,"about_ca_topic_score_gemma":0.00008970776,"domain_scores_codex":[0.9966128,0.0006445333,0.0005406808,0.0002323927,0.001412111,0.0005575194],"domain_scores_gemma":[0.9975778,0.001423401,0.0002526053,0.0003251044,0.0003454971,0.00007557452],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"qualitative","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.001064662,0.0002530738,0.000001223746,0.00007904599,0.00008394803,3.585319e-7,0.8815306,0.007600523,0.0001284242,0.01075341,0.000308133,0.09819658],"study_design_scores_gemma":[0.003224013,0.002642486,0.0005762958,0.00001399507,0.000004985576,0.000001205341,0.1346366,0.8561706,0.0006968114,0.001659139,0.0002264084,0.0001474731],"study_design_candidate":"qualitative","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3642357,0.00005447021,0.6284702,0.001447647,0.0004534081,0.004361255,0.0001040249,0.00009673569,0.0007765497],"genre_scores_gemma":[0.9978529,0.000004407261,0.000349178,0.0000479192,0.00005297416,0.001651535,0.00001924956,0.000007811374,0.00001403256],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.84857,"threshold_uncertainty_score":0.9999384,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1444257212986051,"score_gpt":0.4327990415068596,"score_spread":0.2883733202082545,"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."}}