{"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/hierarchy-aware-biased-bound-margin-loss","title":"Hierarchy-aware Biased Bound Margin Loss Function for Hierarchical Text Classification","arxiv_id":null,"date":"2024-08-13","proceeding":"Findings of the Association for Computational Linguistics ACL 2024 8","authors":["Gibaeg Kim","SangHun Im","Heung-Seon Oh"],"abstract":"Hierarchical text classification (HTC) is a challenging problem with two key issues: utilizing structural information and mitigating label imbalance. Recently, the unit-based approach generating unit-based feature representations has outperformed the global approach focusing on a global feature representation. Nevertheless, unit-based models using BCE and ZLPR losses still face static thresholding and label imbalance challenges. Those challenges become more critical in large-scale hierarchies. This paper introduces a novel hierarchy-aware loss function for unit-based HTC models: Hierarchy-aware Biased Bound Margin (HBM) loss. HBM integrates learnable bounds, biases, and a margin to address static thresholding and mitigate label imbalance adaptively. Experimental results on benchmark datasets demonstrate the superior performance of HBM compared to competitive HTC models.","url_abs":"https://aclanthology.org/2024.findings-acl.457/","url_pdf":"https://aclanthology.org/2024.findings-acl.457.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":"hierarchy-aware-biased-bound-margin-loss","repo_url":"https://github.com/whitepurple/HBM-loss-for-HTC","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"hierarchical-multi-label-classification","task_name":"Hierarchical Multi-label Classification"},{"task_slug":"multi-label-classification","task_name":"Multi-Label Classification"},{"task_slug":"text-classification","task_name":"Text Classification"},{"task_slug":"text-classification-1","task_name":"text-classification"}],"methods":[{"method_slug":"hbm-loss","method_name":"HBM Loss"},{"method_slug":"zlpr-loss","method_name":"ZLPR Loss"}],"datasets_introduced":[],"methods_introduced":[{"slug":"hbm-loss","name":"HBM Loss","full_name":"Hierarchy-aware Biased Bound Margin Loss"}],"results":[{"leaderboard":"/sota/hierarchical-multi-label-classification-on-19","task":"Hierarchical Multi-label Classification","dataset":"EURLEX57K","model":"HiDEC+HBM Loss","rank_in_archive_order":1,"of":4,"metrics":{"Macro F1":"28.77±0.11","Micro F1":"76.48±0.12"},"uses_additional_data":false},{"leaderboard":"/sota/hierarchical-multi-label-classification-on-19","task":"Hierarchical Multi-label Classification","dataset":"EURLEX57K","model":"HPT+HBM Loss","rank_in_archive_order":2,"of":4,"metrics":{"Macro F1":"28.70±0.22","Micro F1":"75.78±0.15"},"uses_additional_data":false},{"leaderboard":"/sota/hierarchical-multi-label-classification-on-19","task":"Hierarchical Multi-label Classification","dataset":"EURLEX57K","model":"HPT","rank_in_archive_order":3,"of":4,"metrics":{"Macro F1":"28.46±0.26","Micro F1":"75.54±0.20"},"uses_additional_data":false},{"leaderboard":"/sota/hierarchical-multi-label-classification-on-19","task":"Hierarchical Multi-label Classification","dataset":"EURLEX57K","model":"HiDEC","rank_in_archive_order":4,"of":4,"metrics":{"Macro F1":"27.91±0.11","Micro F1":"75.14±0.19"},"uses_additional_data":false},{"leaderboard":"/sota/hierarchical-multi-label-classification-on-18","task":"Hierarchical Multi-label Classification","dataset":"New York Times Annotated Corpus","model":"HiDEC+HBM Loss","rank_in_archive_order":1,"of":7,"metrics":{"Macro F1":"70.69±0.19","Micro F1":"80.52±0.18"},"uses_additional_data":false},{"leaderboard":"/sota/hierarchical-multi-label-classification-on-18","task":"Hierarchical Multi-label Classification","dataset":"New York Times Annotated Corpus","model":"HPT+HBM Loss","rank_in_archive_order":2,"of":7,"metrics":{"Macro F1":"70.23±0.18","Micro F1":"80.42±0.12"},"uses_additional_data":false},{"leaderboard":"/sota/hierarchical-multi-label-classification-on-18","task":"Hierarchical Multi-label Classification","dataset":"New York Times Annotated Corpus","model":"HiDEC","rank_in_archive_order":4,"of":7,"metrics":{"Macro F1":"69.80±0.24","Micro F1":"80.13±0.16"},"uses_additional_data":false},{"leaderboard":"/sota/hierarchical-multi-label-classification-on-18","task":"Hierarchical Multi-label Classification","dataset":"New York Times Annotated Corpus","model":"HPT","rank_in_archive_order":5,"of":7,"metrics":{"Macro F1":"69.69±0.49","Micro F1":"80.04±0.23"},"uses_additional_data":false},{"leaderboard":"/sota/hierarchical-multi-label-classification-on-17","task":"Hierarchical Multi-label Classification","dataset":"RCV1-v2","model":"HiDEC+HBM Loss","rank_in_archive_order":1,"of":7,"metrics":{"Macro F1":"71.47±0.20","Micro F1":"87.81±0.09"},"uses_additional_data":false},{"leaderboard":"/sota/hierarchical-multi-label-classification-on-17","task":"Hierarchical Multi-label Classification","dataset":"RCV1-v2","model":"HiDEC","rank_in_archive_order":2,"of":7,"metrics":{"Macro F1":"70.82±0.20","Micro F1":"87.70±0.12"},"uses_additional_data":false},{"leaderboard":"/sota/hierarchical-multi-label-classification-on-17","task":"Hierarchical Multi-label Classification","dataset":"RCV1-v2","model":"HPT+HBM Loss","rank_in_archive_order":3,"of":7,"metrics":{"Macro F1":"70.55±0.13","Micro F1":"87.82±0.06"},"uses_additional_data":false},{"leaderboard":"/sota/hierarchical-multi-label-classification-on-17","task":"Hierarchical Multi-label Classification","dataset":"RCV1-v2","model":"HPT","rank_in_archive_order":4,"of":7,"metrics":{"Macro F1":"70.23±0.31","Micro F1":"87.82±0.14"},"uses_additional_data":false}],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}