{"id":"W3087820244","doi":"10.1145/3395027.3419587","title":"A Framework to Evaluate Webpage Segment Recognizers","year":2020,"lang":"en","type":"article","venue":"","topic":"Web Data Mining and Analysis","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Dalhousie University","funders":"","keywords":"Computer science; Web page; Information retrieval; Search engine; World Wide Web; Natural language processing; Artificial intelligence","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":[0.007118591,0.002003551,0.001355483,0.005938558,0.0009042115,0.003814749,0.00229152,0.002374424,0.004534863],"category_scores_gemma":[0.01606872,0.0004654355,0.001142096,0.001456734,0.001112399,0.003331051,0.002120875,0.001178794,0.002812886],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001843409,"about_ca_system_score_gemma":0.002079858,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008934129,"about_ca_topic_score_gemma":0.005601414,"domain_scores_codex":[0.9949576,0.001037507,0.0005349591,0.0007278552,0.002286857,0.0004552587],"domain_scores_gemma":[0.9929358,0.002219173,0.0005895814,0.0006155264,0.003176213,0.0004635893],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.002263143,0.00247049,0.02861826,0.0008199043,0.0006331704,0.0005709492,0.0004760274,0.1761403,0.1007861,0.03250766,0.01563451,0.6390796],"study_design_scores_gemma":[0.0001459348,0.001545331,0.005265965,0.00009260697,0.0001282488,0.0002201093,0.0002485667,0.9431938,0.03048247,0.0109229,0.007665076,0.00008893899],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1056317,0.0008601739,0.8649439,0.0003018569,0.0001755651,0.00220111,0.002041381,0.01383659,0.01000771],"genre_scores_gemma":[0.4608882,0.0002625546,0.5249556,0.000214934,0.00007937296,0.001727891,0.00504263,0.0006024425,0.006226319],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.008934129,"threshold_uncertainty_score":0.03764719,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05614677183258542,"score_gpt":0.2995525377857658,"score_spread":0.2434057659531804,"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."}}