cs.CL提交于 2026-10-26已译

面向上下文依存句法分析的 POS 签名示范检索

POS-Signature Demonstration Retrieval for In-Context Dependency Parsing

Muhammad Azam Afridi · Rajab Ali · Marco Passarotti

中文摘要

上下文学习的示范通常按语义相似度检索，但该指标未必适合结构化预测。研究面向 Universal Dependencies 句法分析，采用词性签名(POS-signature)策略，以加权词性统计表示句子，并从带标注训练集中检索 3 个长度匹配、已人工分析的样例。在 UD English Web Treebank 测试集全部 2,077 个句子上评测两个开放权重语言模型，不进行微调，并与固定样例、语义检索的三样例基线以及零样例、思维链和批判—改进系统比较。POS 签名检索在两个模型上均取得最高的带标注依存弧得分，分别为 72.4 和 73.2，比语义检索高 1.9 和 2.7 分；两项对比的配对 bootstrap 95% 置信区间均不包含零。思维链与改进策略的比较因模型而异。探索性结果表明，在 English-EWT 上，所测 POS 签名检索方法表现更优。提示词、归档输出和评测代码均已公开。

关键要点

01语义相似度不一定能代表结构化预测中的相关性。
02以加权词性统计构建 POS 签名，检索 3 个长度兼容的标注样例。
03在 English-EWT 全部 2,077 个测试句上评测两个开放权重语言模型。
04POS 签名检索取得 72.4 和 73.2 的最高带标注依存弧得分。
05结果限于 English-EWT 的探索性评测，作者公开提示词、输出与代码。

解读

尚无解读。

原始英文摘要

arXiv:2610.31966v1 Announce Type: new Abstract: Demonstrations for in-context learning are often retrieved by semantic similarity, which may be a poor proxy for relevance in structured prediction. We study demonstration selection for Universal Dependencies parsing with a POS-signature policy that represents sentences with weighted part-of-speech statistics and retrieves three length-compatible, gold-parsed examples from the labeled training split. We evaluate two open-weight language models without fine-tuning on all 2,077 sentences of the UD English Web Treebank test set, comparing the policy with shot- and format-matched fixed and semantic three-shot baselines, as well as zero-shot, chain-of-thought, and critique-refinement systems. POS-signature retrieval yields the highest labeled attachment score among the tested prompting systems for both models (72.4 and 73.2), outperforming semantic retrieval by 1.9 and 2.7 points; paired-bootstrap 95% confidence intervals exclude zero for both contrasts. Chain-of-thought and refinement comparisons vary across models. These exploratory results favor the tested POS-signature retrieval approach on English-EWT. We release the prompts, archived outputs, and evaluation code.