{"id":"W3143367551","doi":"10.21437/interspeech.2021-941","title":"ECAPA-TDNN Embeddings for Speaker Diarization","year":2021,"lang":"en","type":"article","venue":"","topic":"Speech Recognition and Synthesis","field":"Computer Science","cited_by":92,"is_retracted":false,"has_abstract":true,"ca_institutions":"Université de Sherbrooke; Mila - Quebec Artificial Intelligence Institute","funders":"","keywords":"Speaker diarisation; Computer science; Speaker recognition; Robustness (evolution); Speech recognition; Artificial neural network; Discriminative model; Artificial intelligence; Time delay neural network; 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":[0.00007662846,0.00005258124,0.00006802737,0.0000354915,0.00005852559,0.0001430916,0.0001535635,0.00003245613,0.0005129673],"category_scores_gemma":[0.0001414954,0.0000481044,0.00005722692,0.0001914883,0.000007426982,0.0002450805,0.0000487028,0.00002220497,0.00017224],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00001208723,"about_ca_system_score_gemma":0.00003643965,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000002019528,"about_ca_topic_score_gemma":0.000005972178,"domain_scores_codex":[0.9994611,0.00001691968,0.00009764307,0.0002067407,0.00009836752,0.0001191878],"domain_scores_gemma":[0.9995005,0.0001057561,0.0000228203,0.0001890685,0.000135011,0.00004685087],"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":[0.00000281791,0.00009134694,0.0001451623,0.00001492734,0.00002175745,0.00001521268,0.0003037666,0.000003617208,0.01345244,0.3815684,0.02094325,0.5834373],"study_design_scores_gemma":[0.0005593571,0.00003214106,0.001719999,0.00001961642,0.0000123553,0.0000471144,0.000128912,0.1489896,0.6084012,0.03340532,0.2063433,0.000341011],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.001784296,0.00001670563,0.9361389,0.001754287,0.0002490875,0.00006678385,0.000001639522,0.0001323568,0.05985593],"genre_scores_gemma":[0.09964024,0.00001897804,0.8740166,0.003927133,0.0001377905,0.00003112477,0.00001743974,0.00001142269,0.02219925],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.5949488,"threshold_uncertainty_score":0.5616632,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01856822171074096,"score_gpt":0.2489785712134402,"score_spread":0.2304103495026992,"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."}}