{"id":"W2786731957","doi":"10.18653/v1/w17-5532","title":"Generating and Evaluating Summaries for Partial Email Threads: Conversational Bayesian Surprise and Silver Standards","year":2017,"lang":"en","type":"article","venue":"","topic":"Advanced Text Analysis Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"","keywords":"Computer science; Automatic summarization; Surprise; Thread (computing); Annotation; Redundancy (engineering); Information retrieval; Bayesian probability; Natural language processing; Artificial intelligence; Machine learning; Programming language","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":[0.00724354,0.001410205,0.001663729,0.003747034,0.001474384,0.003109115,0.00195938,0.002382489,0.002805203],"category_scores_gemma":[0.0476784,0.0005856736,0.001060815,0.002454693,0.0007699035,0.005383409,0.002869496,0.002163905,0.002400693],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001380037,"about_ca_system_score_gemma":0.002112292,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003068378,"about_ca_topic_score_gemma":0.006354412,"domain_scores_codex":[0.9943879,0.00232409,0.0005502873,0.001439271,0.001038007,0.0002603973],"domain_scores_gemma":[0.9753288,0.01191156,0.00211251,0.004523315,0.005219817,0.0009040222],"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":[0.001483437,0.000330114,0.01197698,0.0009329835,0.0002743666,0.0003859538,0.003049405,0.02981195,0.01841316,0.01958476,0.03630527,0.8774515],"study_design_scores_gemma":[0.0001638804,0.0006284976,0.007103436,0.0002132037,0.0002497172,0.0005017476,0.001204106,0.8349401,0.02935668,0.08005232,0.04543042,0.0001558615],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.06356364,0.001986589,0.9178174,0.001037443,0.0002880631,0.0003218571,0.002958708,0.0092846,0.002741733],"genre_scores_gemma":[0.3192255,0.0004816623,0.6610881,0.0003989602,0.0004886414,0.0005166376,0.01229306,0.001105149,0.004402206],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.00724354,"threshold_uncertainty_score":0.03830796,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03591123173986498,"score_gpt":0.3603061999791383,"score_spread":0.3243949682392733,"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."}}