arXiv cs.CL· Zenith Biswas, Andrew Thomas Dyer·· 12 小时前AI 评分20
质量优于数量：重新审视孟加拉语指代消解的数据与模型假设
Quality over Quantity: Rethinking Data and Model Assumptions for Bengali Coreference Resolution
AI 导读
研究在孟加拉语指代消解任务上评估多语言 Transformer 编码器，发现少量高质量标注数据的表现显著优于更大规模的混合质量数据集（p < 0.01）。回译等数据增强未带来提升甚至可能降低效果，从英语迁移会损害性能、从印地语迁移也无一致收益，且多个常用多语言模型出现嵌入空间坍缩，不适合基于相似度的聚类任务。
正文
Abstract:Multilingual transformer models are often assumed to generalise across languages, but their effectiveness for coreference resolution in low-resource settings remains unverified. We investigate this gap in Bengali, a language with over 250 million speakers but limited annotated coreference resources. We evaluate transformer-based pre-trained encoders under multiple settings, including zero-shot, fine-tuning, data augmentation, and cross-lingual transfer. Our results challenge several common assumptions about model generalisation and data efficiency in low-resource NLP. We demonstrate that a small amount of high-quality annotated data significantly outperforms larger mixed-quality datasets (p < 0.01). We find that data augmentation, particularly back-translation, does not improve performance and can degrade results in coreference resolution. Cross-lingual transfer is also inconsistent: transfer from a distant language like English reduces performance, and transfer from the related language Hindi provides no consistent benefit across models. Several widely used multilingual models exhibit collapsed embedding spaces, rendering them unsuitable for similarity-based clustering tasks. These findings highlight the importance of annotation quality over data quantity and provide empirical guidance for developing coreference resolution systems in low-resource languages.
| Comments: | Accepted at KONVENS 2026, Hamburg, Germany. 11 pages, 1 figure, 5 tables. Code: this https URL |
| Subjects: | Computation and Language (cs.CL) |
| ACM classes: | I.2.7 |
| Cite as: | arXiv:2610.35114 [cs.CL] |
| (or arXiv:2610.35114v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.35114 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zenith Biswas [view email]
[v1]
Mon, 26 Oct 2026 09:53:02 UTC (49 KB)
来源：arXiv cs.CL · arxiv.org