{"id":"W4389523795","doi":"10.18653/v1/2023.emnlp-main.11","title":"Better Quality Pre-training Data and T5 Models for African Languages","year":2023,"lang":"en","type":"article","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"Natural Sciences and Engineering Research Council of Canada; Compute Canada","keywords":"Computer science; Natural language processing; Quality (philosophy); Training set; Training (meteorology); Artificial intelligence; Philosophy; Geography; Epistemology","routes":{"ca_aff":true,"ca_fund":true,"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":[0.004301969,0.002217225,0.001118992,0.001608441,0.001349362,0.001800206,0.002431718,0.002740543,0.01434518],"category_scores_gemma":[0.01405524,0.000837882,0.001815857,0.001383579,0.0005658438,0.005729519,0.002308884,0.0041452,0.008157163],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001062312,"about_ca_system_score_gemma":0.001827137,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0253319,"about_ca_topic_score_gemma":0.03266796,"domain_scores_codex":[0.9984044,0.0007593789,0.0001296198,0.0003652807,0.0001279107,0.000213369],"domain_scores_gemma":[0.9946511,0.002844249,0.0001387899,0.000936891,0.00120072,0.0002281795],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.004797678,0.00110608,0.01817058,0.0009123211,0.0004821238,0.0006562431,0.0005600964,0.1458779,0.01408433,0.00598009,0.1266726,0.6807],"study_design_scores_gemma":[0.000518542,0.000410283,0.006358861,0.0003354104,0.0002451484,0.0002145386,0.001032994,0.9418287,0.01709579,0.01087041,0.02097774,0.0001115189],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.5024521,0.0119607,0.393058,0.009131894,0.003525414,0.0006491548,0.02699649,0.03213215,0.02009411],"genre_scores_gemma":[0.6914014,0.00134747,0.2188318,0.001256095,0.0003694532,0.0004468556,0.07151236,0.00278167,0.01205301],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.0253319,"threshold_uncertainty_score":0.05036885,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1201009790835688,"score_gpt":0.3951933630774996,"score_spread":0.2750923839939308,"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."}}