{"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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.04634431,0.002255779,0.001736708,0.004840367,0.002517697,0.003280611,0.002595932,0.002769447,0.001086044],"category_scores_gemma":[0.09129391,0.000772381,0.001933981,0.003686716,0.00291125,0.005179777,0.006593647,0.003322881,0.001281047],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001502976,"about_ca_system_score_gemma":0.001025058,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004229527,"about_ca_topic_score_gemma":0.006987137,"domain_scores_codex":[0.9527655,0.02821201,0.00253611,0.01159818,0.003944731,0.0009434923],"domain_scores_gemma":[0.8737211,0.07272739,0.008604303,0.03820324,0.005705388,0.001038584],"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":[0.002300045,0.001395328,0.2141366,0.0009961093,0.002353572,0.001368746,0.00530867,0.0797985,0.02713406,0.007339108,0.03243756,0.6254317],"study_design_scores_gemma":[0.0001822488,0.000940663,0.1762602,0.0003742816,0.0006774715,0.00255819,0.003763883,0.7138374,0.0402349,0.03753078,0.02332299,0.0003168636],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6337358,0.003511658,0.3429918,0.001663074,0.0004895841,0.000807689,0.003362477,0.007611699,0.005826232],"genre_scores_gemma":[0.8862119,0.0003585346,0.100497,0.0006981582,0.0002124103,0.0004709008,0.009433554,0.0005438745,0.00157363],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.04634431,"threshold_uncertainty_score":0.2450951,"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."}}