{"id":"W3140050815","doi":"10.18653/v1/2021.eacl-main.241","title":"Better Neural Machine Translation by Extracting Linguistic Information from BERT","year":2021,"lang":"en","type":"preprint","venue":"","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":1,"is_retracted":false,"has_abstract":true,"ca_institutions":"Simon Fraser University","funders":"Natural Sciences and Engineering Research Council of Canada; Ministère de la Défense Nationale","keywords":"Computer science; Machine translation; Transformer; Embedding; Natural language processing; Artificial intelligence; Rule-based machine translation; Language model; Point (geometry); Syntax; Semantics (computer science); Programming language; Mathematics; Engineering","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001117689,0.001544762,0.0008643994,0.001100335,0.0004719804,0.001173338,0.0007353433,0.0009927299,0.003648437],"category_scores_gemma":[0.006299948,0.000684518,0.0007697241,0.001305149,0.0006305322,0.00373558,0.001709502,0.002039394,0.003183717],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0005014344,"about_ca_system_score_gemma":0.0007833768,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001812466,"about_ca_topic_score_gemma":0.003767255,"domain_scores_codex":[0.9993054,0.000242403,0.0000453178,0.0002323669,0.0001270758,0.00004753914],"domain_scores_gemma":[0.9984844,0.0006668748,0.0001109731,0.000407798,0.0002892447,0.00004074802],"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":[0.0003705187,0.0002334726,0.001706737,0.0003512406,0.0001386949,0.00035135,0.0002971031,0.2849366,0.05846241,0.02255108,0.01199948,0.6186014],"study_design_scores_gemma":[0.0000282312,0.00008124884,0.000424716,0.00002280842,0.00003218461,0.00007894323,0.00004025767,0.9530414,0.01072087,0.03095211,0.004552958,0.00002428875],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.04905663,0.0006036874,0.9393232,0.0005593247,0.0001844024,0.00006377769,0.0005177651,0.005431223,0.004259947],"genre_scores_gemma":[0.5497285,0.0005929956,0.4375526,0.0004611435,0.0002146071,0.0001730304,0.004058708,0.001221849,0.005996647],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.003648437,"threshold_uncertainty_score":0.01220524,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01349544561993601,"score_gpt":0.2664848690362658,"score_spread":0.2529894234163299,"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."}}