{"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/exploiting-coarse-to-fine-task-transfer-for","title":"Exploiting Coarse-to-Fine Task Transfer for Aspect-level Sentiment Classification","arxiv_id":"1811.10999","date":"2018-11-16","proceeding":"AAAI 2019 2018 11","authors":["Zheng Li","Ying WEI","Yu Zhang","Xiang Zhang","Xin Li","Qiang Yang"],"abstract":"Aspect-level sentiment classification (ASC) aims at identifying sentiment\npolarities towards aspects in a sentence, where the aspect can behave as a\ngeneral Aspect Category (AC) or a specific Aspect Term (AT). However, due to\nthe especially expensive and labor-intensive labeling, existing public corpora\nin AT-level are all relatively small. Meanwhile, most of the previous methods\nrely on complicated structures with given scarce data, which largely limits the\nefficacy of the neural models. In this paper, we exploit a new direction named\ncoarse-to-fine task transfer, which aims to leverage knowledge learned from a\nrich-resource source domain of the coarse-grained AC task, which is more easily\naccessible, to improve the learning in a low-resource target domain of the\nfine-grained AT task. To resolve both the aspect granularity inconsistency and\nfeature mismatch between domains, we propose a Multi-Granularity Alignment\nNetwork (MGAN). In MGAN, a novel Coarse2Fine attention guided by an auxiliary\ntask can help the AC task modeling at the same fine-grained level with the AT\ntask. To alleviate the feature false alignment, a contrastive feature alignment\nmethod is adopted to align aspect-specific feature representations\nsemantically. In addition, a large-scale multi-domain dataset for the AC task\nis provided. Empirically, extensive experiments demonstrate the effectiveness\nof the MGAN.","url_abs":"http://arxiv.org/abs/1811.10999v1","url_pdf":"http://arxiv.org/pdf/1811.10999v1.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":"exploiting-coarse-to-fine-task-transfer-for","repo_url":"https://github.com/hsqmlzno1/MGAN","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"sentence","task_name":"Sentence"},{"task_slug":"sentiment-analysis","task_name":"Sentiment Analysis"},{"task_slug":"sentiment-classification","task_name":"Sentiment Classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/aspect-based-sentiment-analysis-on-semeval","task":"Aspect-Based Sentiment Analysis (ABSA)","dataset":"SemEval-2014 Task-4","model":"MGAN","rank_in_archive_order":23,"of":48,"metrics":{"Laptop (Acc)":"76.21","Mean Acc (Restaurant + Laptop)":"78.85","Restaurant (Acc)":"81.49"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1811.10999","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}