{"id":"W1965372953","doi":"10.1145/2766462.2767751","title":"Evaluating Streams of Evolving News Events","year":2015,"lang":"en","type":"article","venue":"","topic":"Web Data Mining and Analysis","field":"Computer Science","cited_by":5,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Waterloo","funders":"Networks of Centres of Excellence of Canada; Natural Sciences and Engineering Research Council of Canada","keywords":"Computer science; Automatic summarization; TRACE (psycholinguistics); Event (particle physics); Context (archaeology); Track (disk drive); Reading (process); Information retrieval; Measure (data warehouse); Data mining","routes":{"ca_aff":true,"ca_fund":true,"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":[0.0005221156,0.00004992531,0.0001005996,0.00005845284,0.00002255395,0.0000320506,0.0005074461,0.00001609104,0.00003133867],"category_scores_gemma":[0.0002388692,0.00003973029,0.00003649759,0.0002805596,0.000007343886,0.0003463464,0.0002169735,0.00003128794,0.00004397562],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00001431836,"about_ca_system_score_gemma":0.00007612064,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004132025,"about_ca_topic_score_gemma":0.00002093596,"domain_scores_codex":[0.9992172,0.00004884592,0.0001579403,0.0001692319,0.0002994202,0.0001074338],"domain_scores_gemma":[0.999279,0.00004096467,0.00007176631,0.0004241778,0.000111746,0.00007241022],"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":[0.000003518433,0.0001375406,0.05576565,0.00001463445,0.00008488967,0.000004277482,0.002102513,0.001585674,0.002492154,0.005607224,0.01836001,0.9138419],"study_design_scores_gemma":[0.0002814255,0.00009386009,0.001201208,0.0000229684,0.00001302732,0.000001984224,0.000321084,0.9946648,0.001497911,0.001617083,0.0001906801,0.00009393939],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.3635022,0.0001213812,0.6099933,0.0004395229,0.0001963589,0.00005154308,0.000002226307,0.0001274348,0.02556602],"genre_scores_gemma":[0.8114142,8.779065e-7,0.1878047,0.00004089847,0.00002147145,9.311746e-7,0.000002261797,0.000001974598,0.0007126419],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.9930791,"threshold_uncertainty_score":0.1620154,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1252147845177404,"score_gpt":0.3651689624897769,"score_spread":0.2399541779720366,"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."}}