{"id":"W2900161882","doi":"10.1109/icassp.2019.8682611","title":"Adapting End-to-end Neural Speaker Verification to New Languages and Recording Conditions with Adversarial Training","year":2019,"lang":"en","type":"preprint","venue":"","topic":"Speech Recognition and Synthesis","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Computer Research Institute of Montréal; McGill University","funders":"","keywords":"Adversarial system; Computer science; NIST; Speaker verification; Speech recognition; Task (project management); Artificial neural network; Speaker recognition; Embedding; End-to-end principle; Margin (machine learning); Artificial intelligence; Training (meteorology); Speaker diarisation; Residual; Combing; Machine learning; Algorithm; Engineering","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":[0.0002704019,0.0002346906,0.0002874123,0.0002753575,0.0001132172,0.0004480945,0.0004366433,0.0001163841,0.0004321712],"category_scores_gemma":[0.0001417031,0.0002120943,0.00006174111,0.0002205146,0.00001918484,0.0002715975,0.0003934454,0.0002662882,0.0001384166],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00005507815,"about_ca_system_score_gemma":0.0001850354,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0007452517,"about_ca_topic_score_gemma":0.0002557694,"domain_scores_codex":[0.9983792,0.00007496181,0.0002508084,0.0007456705,0.0002726994,0.0002766042],"domain_scores_gemma":[0.9988026,0.0002229138,0.0001234176,0.0005055362,0.00007846286,0.000267069],"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":[0.00004361928,0.00002447862,0.0004354088,0.00003683602,0.00009296699,0.00002459322,0.008017832,0.001560212,0.002151789,0.004047764,0.003395428,0.9801691],"study_design_scores_gemma":[0.01024201,0.002682529,0.1150695,0.006260287,0.0009938007,0.001593883,0.05223934,0.6482762,0.03833453,0.0101355,0.1002784,0.01389408],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1286321,0.00002256212,0.8396264,0.005396799,0.001008888,0.0007806809,0.00003590946,0.0003911362,0.02410547],"genre_scores_gemma":[0.6700161,0.000004146727,0.3270962,0.0009165849,0.0002569427,0.00002825376,0.00003228221,0.00001887169,0.001630608],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.966275,"threshold_uncertainty_score":0.8648954,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05388103435155326,"score_gpt":0.2965635239524541,"score_spread":0.2426824896009008,"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."}}