{"id":"W2091181450","doi":"10.1117/12.476317","title":"Incorporating Audio Cues into Dialog and Action Scene Extraction","year":2003,"lang":"en","type":"article","venue":"Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE","topic":"Video Analysis and Summarization","field":"Computer Science","cited_by":28,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"","keywords":"Dialog box; Computer science; Audio visual; Classifier (UML); Artificial intelligence; Speech recognition; Action recognition; Action (physics); Audio signal processing; Feature extraction; Computer vision; Audio signal; Multimedia; Speech coding; World Wide Web","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":[0.0004488411,0.001667807,0.0009313773,0.002483402,0.0007243206,0.001300934,0.0008502278,0.0008137112,0.003042498],"category_scores_gemma":[0.001706708,0.0004344843,0.0007929655,0.0008109543,0.0003660563,0.001636247,0.0008514273,0.0009480565,0.001996782],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003903896,"about_ca_system_score_gemma":0.0007137824,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003571666,"about_ca_topic_score_gemma":0.008536227,"domain_scores_codex":[0.9994967,0.00008083377,0.00003376411,0.0001624025,0.0001343721,0.00009195648],"domain_scores_gemma":[0.9991517,0.0003249983,0.00009129867,0.0000744969,0.000279368,0.00007814195],"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":[0.0005316235,0.00025577,0.002694629,0.0006709637,0.00008615216,0.0004565449,0.0003590776,0.007603744,0.2867379,0.002800674,0.006459229,0.6913438],"study_design_scores_gemma":[0.00009232974,0.0009282331,0.02735976,0.0002454811,0.0003549276,0.001241911,0.001844088,0.4247323,0.4718667,0.01764127,0.05345482,0.0002382147],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.05231109,0.001144488,0.9278353,0.0002566503,0.000180401,0.0005804735,0.002163024,0.008837636,0.006691026],"genre_scores_gemma":[0.2221203,0.0007210621,0.7653705,0.0001889017,0.0002008692,0.0003346015,0.005351627,0.0004876415,0.005224445],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.003571666,"threshold_uncertainty_score":0.01017815,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.0139327860538024,"score_gpt":0.2414906279017053,"score_spread":0.2275578418479029,"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."}}