{"id":"W1973458428","doi":"10.1023/b:joms.0000021524.90614.a9","title":"A New Look at Information Retrieval Evaluation: Proposal for Solutions","year":2004,"lang":"en","type":"article","venue":"Journal of Medical Systems","topic":"Information Retrieval and Search Behavior","field":"Computer Science","cited_by":13,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of Victoria","funders":"Medical Research Council","keywords":"Relevance (law); Computer science; Terminology; Information retrieval; Identification (biology); Precision and recall; Information system; Data science; Management science; Data mining","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":[0.005462148,0.00008863975,0.000200071,0.0001841042,0.000198112,0.0001836737,0.0007353878,0.0001499737,0.0000719406],"category_scores_gemma":[0.001069735,0.00006264821,0.00013435,0.0003324458,0.00003281964,0.001953894,0.0001176055,0.0002333026,0.000091759],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0005350393,"about_ca_system_score_gemma":0.004433207,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00002948623,"about_ca_topic_score_gemma":0.000003321326,"domain_scores_codex":[0.9950137,0.00008091004,0.000939442,0.00006851413,0.003640688,0.0002567479],"domain_scores_gemma":[0.9976512,0.00008712772,0.0005263005,0.0001662206,0.001102192,0.0004669943],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0006382965,0.000351914,0.0002167713,0.0006518634,0.0002274579,0.00006177178,0.01025075,0.008691551,0.0004340392,0.6126062,0.1784095,0.1874599],"study_design_scores_gemma":[0.03536627,0.004757441,0.002598025,0.002077015,0.0002516315,0.008011322,0.001394774,0.5364746,0.002379874,0.02235186,0.3833014,0.001035827],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.005017075,0.0002602219,0.9775737,0.01398167,0.002320466,0.0006142774,0.000006000385,0.00002807365,0.0001985177],"genre_scores_gemma":[0.9028766,0.0001644794,0.0895803,0.001418639,0.004486533,0.00006369401,0.00005037746,0.000025622,0.001333769],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.8978595,"threshold_uncertainty_score":0.7864319,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04250847837540634,"score_gpt":0.3186410705850684,"score_spread":0.2761325922096621,"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."}}