{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.0002691304,0.0004037829,0.0003705115,0.0002850161,0.0001331328,0.0004613801,0.001718721,0.0004216173,0.00003040844],"category_scores_gemma":[0.00003377566,0.000361958,0.0001900374,0.0002886575,0.00003238787,0.0006081074,0.001036197,0.001229284,0.00002877218],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001118026,"about_ca_system_score_gemma":0.0001267125,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0005455597,"about_ca_topic_score_gemma":0.00006493898,"domain_scores_codex":[0.9978507,0.00007600674,0.0004741898,0.0008817252,0.0003520182,0.0003654016],"domain_scores_gemma":[0.9981199,0.0001172742,0.0002962051,0.001224115,0.0001496011,0.00009294251],"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":[0.00002879052,0.0001429892,0.0006931731,0.0004874346,0.00009652587,0.00003799456,0.001814363,0.006062691,0.01737101,0.010448,0.0005887968,0.9622282],"study_design_scores_gemma":[0.0001861078,0.00004396408,0.00006905934,0.00007294364,0.00002277918,0.00002781565,0.000001719171,0.9722586,0.003901182,0.02281913,0.0001297236,0.0004669733],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0005778688,0.005626148,0.9869391,0.0008829322,0.001471694,0.0007662787,0.00001122872,0.002208232,0.001516493],"genre_scores_gemma":[0.5806357,0.00001299907,0.4185078,0.0001736685,0.00006782215,0.00003061423,0.00006166028,0.00002371099,0.0004860348],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9661959,"threshold_uncertainty_score":0.9998832,"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."}}