{"id":"W1997539970","doi":"10.1109/icassp.2010.5494926","title":"Large margin estimation of n-gram language models for speech recognition via linear programming","year":2010,"lang":"en","type":"article","venue":"","topic":"Speech Recognition and Synthesis","field":"Computer Science","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"York University","funders":"","keywords":"Discriminative model; Computer science; Margin (machine learning); Speech recognition; Language model; Viterbi algorithm; Vocabulary; Word error rate; Viterbi decoder; n-gram; Artificial intelligence; Pattern recognition (psychology); Decoding methods; Hidden Markov model; Algorithm; Machine learning","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":[0.0004882466,0.0001006867,0.0001385487,0.0001303973,0.00006569629,0.00005968691,0.0002576471,0.00009426794,0.0001300163],"category_scores_gemma":[0.0001349704,0.00009098181,0.00009244463,0.0002109329,0.0000188253,0.0005468823,0.0000472362,0.00009970615,0.00005900509],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000008346093,"about_ca_system_score_gemma":0.00002465335,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00003164871,"about_ca_topic_score_gemma":0.00008644178,"domain_scores_codex":[0.9991019,0.00002315308,0.0002383105,0.0002383345,0.0001819347,0.0002164177],"domain_scores_gemma":[0.9992645,0.000129611,0.0001004333,0.0002733609,0.0001669063,0.00006515903],"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":[0.000005884656,0.0001176705,0.000004738744,0.00002709369,0.000007177674,0.000001551569,0.0002255771,0.000008767981,0.004336802,0.002758916,0.00008096178,0.9924248],"study_design_scores_gemma":[0.0003273074,0.00004776821,0.00001607201,0.00001403461,0.000008272587,0.00001491921,0.00007087769,0.8225347,0.163912,0.0124397,0.000492342,0.0001220326],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.05554603,0.00000528548,0.9413152,0.000204055,0.0001907137,0.0004369789,0.00001535421,0.0002136457,0.002072762],"genre_scores_gemma":[0.2581202,0.000001240278,0.7414831,0.00008848274,0.00004498357,0.00005439622,0.0000371297,0.000008326787,0.000162209],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.9923028,"threshold_uncertainty_score":0.3710131,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02641845991397187,"score_gpt":0.2823456849418688,"score_spread":0.255927225027897,"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."}}