{"id":"W4400528580","doi":"10.1145/3626772.3657884","title":"CIRAL: A Test Collection for CLIR Evaluations in African Languages","year":2024,"lang":"en","type":"article","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":2,"is_retracted":false,"has_abstract":true,"ca_institutions":"Huawei Technologies (Canada); University of Waterloo","funders":"Universitas Brawijaya","keywords":"Computer science; Natural language processing; Test (biology); Artificial intelligence; Information retrieval","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":[0.00834597,0.001994657,0.001249408,0.006762973,0.003087886,0.002108127,0.002998533,0.00181855,0.01470095],"category_scores_gemma":[0.03006111,0.0006549247,0.001055714,0.004471567,0.001853097,0.004026591,0.004538723,0.00192728,0.01323415],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001456468,"about_ca_system_score_gemma":0.002992304,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.009930022,"about_ca_topic_score_gemma":0.01517213,"domain_scores_codex":[0.9908729,0.004632787,0.0009696229,0.001047245,0.001976716,0.0005007159],"domain_scores_gemma":[0.9753642,0.009530596,0.0008899773,0.004725643,0.00813456,0.001355061],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.002730757,0.002788687,0.008948996,0.009326117,0.0002420995,0.001762035,0.007117998,0.003343219,0.04257638,0.004057298,0.6233329,0.2937734],"study_design_scores_gemma":[0.002241351,0.003223597,0.07499411,0.001603236,0.0004144124,0.00474439,0.009321927,0.02475069,0.06420666,0.003390311,0.8103575,0.0007518885],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"empirical","genre_gemma":"dataset","genre_scores_codex":[0.4579515,0.01297416,0.0809793,0.003556974,0.002618295,0.02409026,0.2669912,0.05400483,0.09683345],"genre_scores_gemma":[0.2765658,0.001755957,0.1736336,0.001572323,0.0006692875,0.01329609,0.5017288,0.006255684,0.02452242],"genre_candidate":"dataset","genre_consensus":null,"teacher_disagreement_score":0.01470095,"threshold_uncertainty_score":0.04917955,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02130701428789182,"score_gpt":0.3627427712237322,"score_spread":0.3414357569358403,"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."}}