{"id":"W3156463632","doi":"10.1145/3442381.3449977","title":"Typing Errors in Factual Knowledge Graphs: Severity and Possible Ways Out","year":2021,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Alberta","funders":"","keywords":"Computer science; Benchmark (surveying); Code (set theory); Artificial intelligence; Typing; Machine learning; Word error rate; Quality (philosophy); Natural language processing; Noisy data; Programming language; Speech recognition","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":[0.01562773,0.001108369,0.001184803,0.005742567,0.001847397,0.005004442,0.002512744,0.002455888,0.002208506],"category_scores_gemma":[0.188329,0.001107031,0.001011976,0.005404462,0.002699445,0.009648896,0.003223183,0.003040265,0.001033659],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001429678,"about_ca_system_score_gemma":0.001787155,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.006370222,"about_ca_topic_score_gemma":0.01053306,"domain_scores_codex":[0.9802179,0.007428538,0.002050044,0.004067284,0.005274043,0.0009621396],"domain_scores_gemma":[0.7402746,0.1911527,0.01665515,0.03111629,0.01876458,0.002036653],"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":[0.001206498,0.0005518663,0.2953258,0.002487684,0.0006767262,0.002459531,0.01287384,0.0833008,0.01156064,0.06299704,0.04705109,0.4795085],"study_design_scores_gemma":[0.0001007304,0.0002335965,0.058992,0.001450274,0.0005184645,0.004266577,0.007175488,0.6377707,0.03019048,0.2181907,0.04077581,0.0003351272],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4071983,0.003273064,0.5622968,0.004819508,0.0008029966,0.0003675428,0.005164151,0.01011194,0.005965762],"genre_scores_gemma":[0.7865173,0.0009318226,0.2007253,0.0006784606,0.0001926086,0.0001632696,0.006319423,0.002143656,0.002328293],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01562773,"threshold_uncertainty_score":0.08264834,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05126507019639383,"score_gpt":0.2772535504445526,"score_spread":0.2259884802481587,"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."}}