{"id":"W2952153923","doi":"10.18653/v1/p19-1121","title":"Improved Zero-shot Neural Machine Translation via Ignoring Spurious Correlations","year":2019,"lang":"en","type":"preprint","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":87,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Tencent; Samsung Advanced Institute of Technology; Canadian Institute for Advanced Research; Samsung; Nvidia","keywords":"Machine translation; Computer science; Spurious relationship; Translation (biology); Zero (linguistics); Artificial intelligence; Shot (pellet); Degeneracy (biology); Natural language processing; Language model; Speech recognition; Algorithm; Machine learning","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001885513,0.001101072,0.001331962,0.0008172667,0.0009081638,0.001435915,0.00164153,0.001391489,0.004161032],"category_scores_gemma":[0.008832741,0.0006757826,0.000730362,0.001066106,0.001100858,0.002597222,0.002601124,0.001630099,0.002698575],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0006087297,"about_ca_system_score_gemma":0.001954791,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003878308,"about_ca_topic_score_gemma":0.008113934,"domain_scores_codex":[0.9982559,0.0006201051,0.0000865746,0.000462577,0.0004295818,0.0001451459],"domain_scores_gemma":[0.9968534,0.001394313,0.0001781891,0.0009027105,0.0005671295,0.0001043401],"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":[0.001082089,0.0003410922,0.001889529,0.0004203209,0.0003139841,0.0005856348,0.0004792158,0.2110077,0.04313495,0.05187049,0.01481882,0.6740562],"study_design_scores_gemma":[0.00004497778,0.0001238013,0.0004053162,0.00002700859,0.00005261154,0.0001858074,0.00004175325,0.9574723,0.01294675,0.02571101,0.002959909,0.00002885174],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.05074909,0.000726602,0.9380891,0.0003483524,0.0001797946,0.00004876479,0.0001861788,0.004347478,0.005324516],"genre_scores_gemma":[0.6443759,0.0003811096,0.3390568,0.0004174127,0.0001644067,0.0001207453,0.001570526,0.0008500039,0.01306326],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.004161032,"threshold_uncertainty_score":0.01392001,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.02387670835367649,"score_gpt":0.2818461076736029,"score_spread":0.2579693993199264,"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."}}