arXiv：cs.AI· Kevin Guan, Happy Buzaaba, Christiane Fellbaum·· 10 小时前
低资源语言依存句法分析：Biaffine LSTM 与 Transformer 架构对比评测
Dependency Parsing Across the Resource Spectrum: Evaluating Architectures on High and Low-Resource Languages
AI 导读
研究在十二种类型学多样的语言（含低资源非洲语言）上评测了 Biaffine LSTM、Stack-Pointer Network、AfroXLMR-large 和 RemBERT 四种依存句法分析器。
正文
Abstract:Transformer-based models achieve state-of-the-art dependency parsing for high-resource languages, yet their advantage over simpler architectures in low-resource settings remains poorly understood. We evaluate four parsers---the Biaffine LSTM, Stack-Pointer Network, AfroXLMR-large, and RemBERT---across twelve typologically diverse languages, with a focus on low-resource African languages. We find that the Biaffine LSTM consistently outperforms transformer models in low-resource regimes, with transformers recovering their advantage as training data increases. The crossover falls within a resource range typical of treebanks for under-resourced languages. Morphological complexity (measured via MATTR) emerges as a significant secondary predictor of transformers' relative disadvantage after controlling for corpus size. These results indicate that the Biaffine LSTM may be better suited for syntactic tool development in low-resource regimes until sufficient annotated data is available to leverage the representational capacity of pre-trained transformers.
| Comments: | In Proceedings of the First International Conference on Language Technologies for Low-resource Languages (LaTeLL 2026) |
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2605.02608 [cs.CL] |
| (or arXiv:2605.02608v3 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2605.02608 arXiv-issued DOI via DataCite |
Submission history
From: Kevin Guan [view email]
[v1]
Mon, 4 May 2026 13:55:32 UTC (253 KB)
[v2]
Fri, 7 Aug 2026 17:49:56 UTC (271 KB)
[v3]
Mon, 12 Oct 2026 04:55:12 UTC (271 KB)
来源：arXiv：cs.AI · arxiv.org