{"id":"W2111310810","doi":"10.1145/1148170.1148262","title":"Statistical precision of information retrieval evaluation","year":2006,"lang":"en","type":"article","venue":"","topic":"Advanced Text Analysis Techniques","field":"Computer Science","cited_by":107,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Computer science; Data mining; Concordance; Population; Information retrieval; Confidence interval; Statistics; Degree (music); Statistical hypothesis testing; Data collection; Test (biology); Artificial intelligence; 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":[0.1368204,0.001732465,0.003367085,0.01227101,0.001775468,0.006955544,0.004013717,0.003572714,0.001975494],"category_scores_gemma":[0.5522661,0.001293587,0.002563829,0.01116385,0.004669744,0.009290813,0.004865046,0.004400136,0.001214192],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.003507428,"about_ca_system_score_gemma":0.002702758,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003805264,"about_ca_topic_score_gemma":0.002927595,"domain_scores_codex":[0.8319094,0.08507289,0.0109145,0.01958609,0.05027729,0.002239799],"domain_scores_gemma":[0.3363208,0.5252619,0.02699884,0.08046687,0.02997361,0.0009778818],"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":[0.001197008,0.0002555746,0.09124732,0.002021319,0.002992732,0.0003564498,0.00211375,0.2885578,0.00375896,0.1562261,0.01372923,0.4375438],"study_design_scores_gemma":[0.0001535351,0.0006578689,0.04053127,0.0007153678,0.0006785584,0.0009074734,0.0003244592,0.6826802,0.01284773,0.2501026,0.01001958,0.0003814411],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.04577175,0.006599812,0.9335715,0.000757129,0.0003260046,0.0002645917,0.001223097,0.00292724,0.008558953],"genre_scores_gemma":[0.8223702,0.00139341,0.169208,0.0006164747,0.0007398521,0.0008964523,0.00236134,0.001317466,0.00109674],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.1368204,"threshold_uncertainty_score":0.7235842,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01181805756965918,"score_gpt":0.3080099634055263,"score_spread":0.2961919058358671,"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."}}