{"id":"W3092696693","doi":"10.18653/v1/2020.alw-1.20","title":"On Cross-Dataset Generalization in Automatic Detection of Online Abuse","year":2020,"lang":"en","type":"preprint","venue":"","topic":"Hate Speech and Cyberbullying Detection","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":true,"ca_institutions":"National Research Council Canada","funders":"","keywords":"Generalization; Computer science; Ambiguity; Task (project management); Class (philosophy); Set (abstract data type); Artificial intelligence; Test set; Machine learning; Training set; Natural language processing; Test (biology); Domain (mathematical analysis); Data set; Data mining; Information retrieval; Mathematics","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":[0.0001964832,0.0001980128,0.0002543604,0.0002842877,0.00003671506,0.0001422013,0.0006148918,0.0002213479,0.00002677027],"category_scores_gemma":[0.00008265355,0.000192538,0.00006895471,0.0004241563,0.00002277536,0.0001809396,0.0003113846,0.0003703376,0.00004770441],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00009552317,"about_ca_system_score_gemma":0.0000831128,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004275139,"about_ca_topic_score_gemma":0.0004140517,"domain_scores_codex":[0.9983855,0.000102548,0.0004797694,0.0005497403,0.0003142368,0.0001681302],"domain_scores_gemma":[0.9988761,0.00004019537,0.000246786,0.0007033803,0.00006686446,0.00006670521],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0002189081,0.002312924,0.001480003,0.002978191,0.0002554164,0.0002176011,0.004928676,0.3930573,0.1077396,0.01835607,0.005359615,0.4630956],"study_design_scores_gemma":[0.0003757646,0.000178006,0.005558547,0.0001373001,0.000007321949,0.000005000485,0.000005425321,0.9317498,0.0555908,0.005784603,0.0003433023,0.0002640825],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4639865,0.00001866999,0.5343857,0.0001754343,0.0006456252,0.0002956495,0.0001657648,0.0002465647,0.00008009003],"genre_scores_gemma":[0.9767668,0.00004875759,0.02165309,0.0003788873,0.00008339145,0.00002566067,0.0009746357,0.00001675263,0.00005205701],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.5386925,"threshold_uncertainty_score":0.7851471,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02261950507012783,"score_gpt":0.2895182201569829,"score_spread":0.266898715086855,"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."}}