arXiv：cs.CL· Erik Henriksson, Amanda Myntti, Saara Hellstr\"om, Anni Eskelinen, Selcen Erten-Johansson, Veronika Laippala·· 4 小时前AI 评分32
用多语言深度学习自动识别开放网络上的语域
Automatic register identification for the open web using multilingual deep learning
AI 导读
研究提出多语言深度学习模型，可在 16 种语言中识别新闻、论坛等网络语域，并发布含超 72,000 篇文档、覆盖 25 种语域层级分类的 Multilingual CORE 语料库。最佳模型多标签分类平均 F1 达 79%，经数据剪枝去除标签不确定文档后升至 90% 以上，表明性能天花板源于网络语域本身的固有歧义。多语言模型持续优于单语言模型，在未见语言上零样本性能平均下降 7%（3%–8%）。
正文
Abstract:This article presents multilingual deep learning models for identifying web registers -- text varieties such as news reports and discussion forums -- across 16 languages. We introduce the Multilingual CORE corpora, which contain over 72,000 documents annotated with a hierarchical taxonomy of 25 registers designed to cover the entire open web. Using multi-label classification, our best model achieves 79% F1 averaged across languages, matching or exceeding previous studies that used simpler classification schemes. This demonstrates that models can perform well even with a complex register scheme at multilingual scale. However, we observe a consistent performance ceiling across all models and configurations. When we remove documents with uncertain labels through data pruning, performance increases to over 90% F1, suggesting that this ceiling stems from inherent ambiguity in web registers rather than model limitations. Analysis of hybrid texts (those combining multiple registers) reveals that the main challenge lies not in classifying hybrids themselves, but in distinguishing hybrid from non-hybrid documents. Multilingual models consistently outperform monolingual ones, particularly for languages with limited training data. Zero-shot performance on unseen languages drops by an average of 7%, though this varies by language (3--8%), indicating that while registers share features across languages, they also retain language-specific characteristics.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2406.19892 [cs.CL] |
| (or arXiv:2406.19892v5 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2406.19892 arXiv-issued DOI via DataCite |
Submission history
From: Erik Henriksson [view email]
[v1]
Fri, 28 Jun 2024 13:00:30 UTC (13,507 KB)
[v2]
Fri, 19 Jul 2024 20:40:53 UTC (13,505 KB)
[v3]
Tue, 10 Dec 2024 12:46:39 UTC (6,587 KB)
[v4]
Fri, 6 Feb 2026 19:57:26 UTC (2,100 KB)
[v5]
Fri, 2 Oct 2026 07:53:44 UTC (2,077 KB)
来源：arXiv：cs.CL · arxiv.org