{"id":"W4416397568","doi":"10.1016/j.specom.2025.103330","title":"Towards unsupervised speech recognition without pronunciation models","year":2025,"lang":"en","type":"article","venue":"Speech Communication","topic":"Speech Recognition and Synthesis","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"Artificial Intelligence in Medicine (Canada)","funders":"Institute for Information and Communications Technology Promotion","keywords":"Pronunciation; Word (group theory); Pipeline (software); Unsupervised learning; Segmentation; Vocabulary; Word error rate; Speech corpus; Joint (building)","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":[0.001346695,0.001655477,0.001463188,0.0008860074,0.0006551541,0.002121523,0.001655398,0.001967185,0.005234666],"category_scores_gemma":[0.004251966,0.001040435,0.001430884,0.0007507156,0.0007378102,0.002339576,0.002203764,0.00280195,0.01210689],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004268444,"about_ca_system_score_gemma":0.001484286,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003137919,"about_ca_topic_score_gemma":0.005803899,"domain_scores_codex":[0.9984329,0.0004430867,0.0000979027,0.0005131711,0.0003890675,0.000123734],"domain_scores_gemma":[0.9967728,0.00139839,0.0001126954,0.0007580997,0.0008728578,0.00008518793],"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":[0.0003396388,0.0001728493,0.0009656724,0.0003062762,0.0001640626,0.0001727566,0.0001881142,0.03075035,0.1553551,0.01467913,0.008715241,0.7881908],"study_design_scores_gemma":[0.00004334112,0.0001692018,0.001553719,0.00006692796,0.0001436669,0.0004685615,0.0001105795,0.8325859,0.1251815,0.02041745,0.01919761,0.00006160209],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.003543015,0.0002181922,0.990738,0.0001101968,0.0001130619,0.00003282159,0.0002054464,0.003706092,0.001333106],"genre_scores_gemma":[0.09837645,0.0005565032,0.8796084,0.0004461844,0.0002673657,0.0002075288,0.003117009,0.001268972,0.01615169],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.005234666,"threshold_uncertainty_score":0.01751167,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0460206974990346,"score_gpt":0.2815052824270141,"score_spread":0.2354845849279795,"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."}}