{"id":"W2152382718","doi":"10.3115/1626431.1626478","title":"Stabilizing minimum error rate training","year":2009,"lang":"en","type":"article","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":46,"is_retracted":false,"has_abstract":true,"ca_institutions":"National Research Council Canada","funders":"","keywords":"Computer science; Stability (learning theory); Word error rate; Variation (astronomy); Training set; Feature (linguistics); Test data; Component (thermodynamics); Machine translation; Translation (biology); Artificial intelligence; Test (biology); Machine learning; Training (meteorology); Algorithm; Pattern recognition (psychology)","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":[0.0003253582,0.00009832888,0.0001068632,0.00007627101,0.00007624624,0.000168238,0.0006965063,0.00004474105,0.0000186963],"category_scores_gemma":[0.00008054822,0.00007789901,0.00003553186,0.0002911933,0.00001596323,0.0005901814,0.00007406777,0.0001212572,0.00001413375],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00002279816,"about_ca_system_score_gemma":0.00004403833,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000004001828,"about_ca_topic_score_gemma":0.000002124934,"domain_scores_codex":[0.9991831,0.00003325333,0.0001403334,0.0002686677,0.0001275223,0.0002471308],"domain_scores_gemma":[0.9994788,0.00004869402,0.00004449835,0.0003233727,0.00004586654,0.00005876649],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.0000044008,0.00003419087,0.00001932574,0.000007852021,0.000003832626,0.00004566636,0.003426082,0.000003155972,0.08625179,0.2611619,0.001559746,0.647482],"study_design_scores_gemma":[0.0005064827,0.000403056,0.0005786538,0.0001156324,0.000007514368,0.00007832088,0.0004238198,0.06010468,0.2862253,0.6462141,0.004422293,0.0009202002],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01131697,0.0006010131,0.9716021,0.004480163,0.0001102486,0.0001052582,2.209987e-7,0.002062697,0.009721329],"genre_scores_gemma":[0.5270665,9.272581e-7,0.4713164,0.001324589,0.0000211453,0.000001446504,2.615312e-7,0.000002447111,0.0002662721],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.6465619,"threshold_uncertainty_score":0.317663,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03676422810836918,"score_gpt":0.3020566031008958,"score_spread":0.2652923749925267,"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."}}