{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/maximum-bayes-smatch-ensemble-distillation","title":"Maximum Bayes Smatch Ensemble Distillation for AMR Parsing","arxiv_id":"2112.07790","date":"2021-12-14","proceeding":"NAACL 2022 7","authors":["Young-suk Lee","Ramon Fernandez Astudillo","Thanh Lam Hoang","Tahira Naseem","Radu Florian","Salim Roukos"],"abstract":"AMR parsing has experienced an unprecendented increase in performance in the last three years, due to a mixture of effects including architecture improvements and transfer learning. Self-learning techniques have also played a role in pushing performance forward. However, for most recent high performant parsers, the effect of self-learning and silver data augmentation seems to be fading. In this paper we propose to overcome this diminishing returns of silver data by combining Smatch-based ensembling techniques with ensemble distillation. In an extensive experimental setup, we push single model English parser performance to a new state-of-the-art, 85.9 (AMR2.0) and 84.3 (AMR3.0), and return to substantial gains from silver data augmentation. We also attain a new state-of-the-art for cross-lingual AMR parsing for Chinese, German, Italian and Spanish. Finally we explore the impact of the proposed technique on domain adaptation, and show that it can produce gains rivaling those of human annotated data for QALD-9 and achieve a new state-of-the-art for BioAMR.","url_abs":"https://arxiv.org/abs/2112.07790v2","url_pdf":"https://arxiv.org/pdf/2112.07790v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"maximum-bayes-smatch-ensemble-distillation","repo_url":"https://github.com/IBM/transition-amr-parser","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"maximum-bayes-smatch-ensemble-distillation","repo_url":"https://github.com/ibm/amr-annotations","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null},{"paper_slug":"maximum-bayes-smatch-ensemble-distillation","repo_url":"https://github.com/pournaki/transition-amr-parser","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"amr-parsing","task_name":"AMR Parsing"},{"task_slug":"data-augmentation","task_name":"Data Augmentation"},{"task_slug":"domain-adaptation","task_name":"Domain Adaptation"},{"task_slug":"self-learning","task_name":"Self-Learning"},{"task_slug":"transfer-learning","task_name":"Transfer Learning"}],"methods":[{"method_slug":"self-learning","method_name":"Self-Learning"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/amr-parsing-on-bio","task":"AMR Parsing","dataset":"Bio","model":"StructBART + MBSE (IBM)","rank_in_archive_order":1,"of":5,"metrics":{"Smatch":"66.9"},"uses_additional_data":true},{"leaderboard":"/sota/amr-parsing-on-ldc2017t10","task":"AMR Parsing","dataset":"LDC2017T10","model":"StructBART + MBSE (IBM)","rank_in_archive_order":1,"of":27,"metrics":{"Smatch":"86.7"},"uses_additional_data":true},{"leaderboard":"/sota/amr-parsing-on-ldc2017t10","task":"AMR Parsing","dataset":"LDC2017T10","model":"StructBART + MBSE (IBM)","rank_in_archive_order":4,"of":27,"metrics":{"Smatch":"85.9"},"uses_additional_data":true},{"leaderboard":"/sota/amr-parsing-on-ldc2020t02","task":"AMR Parsing","dataset":"LDC2020T02","model":"Graphene Smatch (MBSE paper) (IBM)","rank_in_archive_order":1,"of":13,"metrics":{"Smatch":"85.4"},"uses_additional_data":true},{"leaderboard":"/sota/amr-parsing-on-ldc2020t02","task":"AMR Parsing","dataset":"LDC2020T02","model":"StructBART + MBSE (IBM)","rank_in_archive_order":5,"of":13,"metrics":{"Smatch":"84.3"},"uses_additional_data":true}],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=2112.07790","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}