用渐进微调Transformer实现泰米尔语上下文纠错，显著提升语法一致性与音变规则处理能力。
Contextual Tamil Spelling and Grammar Correction Using Progressively Fine-Tuned Sequence-to-Sequence Transformers
- 分四阶段渐进微调模型，逐步攻克表面错误、语法一致性和跨词音变难题。
- 最佳模型在1000句诊断集上达69.3%精确匹配率，音变纠错准确率达87.5%。
- 首次揭示纠错中召回率与准确率的权衡关系，证明专用监督不可或缺。
泰米尔语拼写与语法纠错因属于黏着型低资源语言，具有丰富的动词形态、复杂的词间音变（sandhi）规则及247个独立字母而极具挑战。以往方法仅处理词级表层错误，难以应对主谓一致、时态一致或跨词音变等需要句级理解的上下文错误。本文提出端到端序列到序列框架，基于合成语料（最多657,720对噪声-清洁句）对mT5-small和mBART-50进行四阶段渐进微调：依次针对表面噪声（v2）、上下文语法（v3）、单点音变（v4）和多点跨词音变（v5）。在1,000句独立验证集上，mBART-50 v5达69.3% top-1精确匹配率，音变正确率达87.5%，主谓一致达43.5%。渐进策略显著提升性能：引入上下文对后主谓一致从1.0%升至52.5%，引入多点音变对后音变准确率从0%升至87.5%。此外，首次量化了精度-召回权衡关系——音变召回率提升以身份准确率下降为代价。Tamil-LLaMA-7B-Instruct在零样本下仅达19.0%，三示例提升至24.7%，仍低于20.0%的复制基线，表明指令模型无法直接迁移到特定句级纠错任务。
原文摘要 · Abstract (English)
Tamil spell and grammar correction is challenging because Tamil is an agglutinative low-resource language with rich verbal morphology, complex sandhi (phonetic transformation) rules at word boundaries, and a script of 247 distinct letters. Prior work targets word-level surface errors with rule-based methods, statistical n-gram models, Minimum Edit Distance, or hybrid pipelines with a transformer re-ranker; such methods cannot reliably handle contextual errors - subject-verb agreement, tense consistency, or cross-word sandhi - which require sentence-level understanding. We propose an end-to-end sequence-to-sequence formulation and fine-tune mT5-small and mBART-50 on a synthetic corpus of up to 657,720 noisy-clean Tamil sentence pairs spanning ten error categories. Both backbones follow the same four-stage progressive schedule, each stage targeting one weakness: surface noise (v2), contextual grammar (v3), single-site sandhi (v4), and multi-site cross-word sandhi (v5). On a 1,000-sentence balanced diagnostic set verified disjoint from all training data, our best model, mBART-50 v5, reaches 69.3% top-1 exact-match accuracy, with 87.5% on sandhi and 43.5% on subject-verb agreement. The schedule is what produces these gains: subject-verb accuracy rises from 1.0% to 52.5% once contextual pairs are introduced, and sandhi from 0% to 87.5% once multi-site sandhi pairs are. We additionally quantify a precision-recall trade-off this literature has not reported: sandhi recall is paid for monotonically in identity accuracy. Finally, Tamil-LLaMA-7B-Instruct reaches 19.0% zero-shot and 24.7% with three demonstrations against a 20.0% copy baseline, showing that a Tamil-adapted instruction model does not transfer to specialised sentence-level correction without task-specific supervision.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。