Scholarlink

<p>Language dialects and registers represent complex data shifts. The use of LLMs can take advantage of more abstract semantic features compared to traditional approaches based on word frequencies. We construct a register-aware benchmark spanning five countries (Argentina, Chile, Colombia, Mexico, Spain) and three registers (formal, informal, mixed). Four open small LLMs (<Emphasis FontCategory="SansSerif">LLaMA</Emphasis>, <Emphasis FontCategory="SansSerif">Gemma</Emphasis>, <Emphasis FontCategory="SansSerif">Mistral</Emphasis>, <Emphasis FontCategory="SansSerif">DeepSeek</Emphasis>) are fine-tuned for 5-way classification, evaluated with all train-test register pairings, and compared with an inexpensive classical TF-IDF + <Emphasis FontCategory="SansSerif">Naïve Bayes</Emphasis> setting. Several main findings emerge: (i) We find a specialization-generalization trade-off and asymmetric cross-register transfer: informal <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(\rightarrow \)</EquationSource> <EquationSource Format="MATHML"><math> <mo stretchy="false">→</mo> </math></EquationSource> </InlineEquation> formal degrades far less than formal <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\rightarrow \)</EquationSource> <EquationSource Format="MATHML"><math> <mo stretchy="false">→</mo> </math></EquationSource> </InlineEquation> informal; (ii) misclassifications are structured: confusions concentrate in linguistically proximate pairs (e.g., Mexico-Colombia) and align with a low-dimensional geometry from Jensen-Shannon distances; (iii) errors stem from two main sources: over-reliance on toponyms/proper nouns (often defaulting to Spain), and interference from shared slang/morphology.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

From formal to informal: how register shift confuses LLMs in Spanish dialect recognition

  • 期刊论文

  • 发表日期：

    2026年6月23日

查看全文
  • Samuel Axel Zander-Marzano,
  • José Hernández-Orallo,
  • Fernando Martínez-Plumed,
  • Pau Amores-Giner,
  • Behzad Mehrbakhsh

摘要

Language dialects and registers represent complex data shifts. The use of LLMs can take advantage of more abstract semantic features compared to traditional approaches based on word frequencies. We construct a register-aware benchmark spanning five countries (Argentina, Chile, Colombia, Mexico, Spain) and three registers (formal, informal, mixed). Four open small LLMs (LLaMA, Gemma, Mistral, DeepSeek) are fine-tuned for 5-way classification, evaluated with all train-test register pairings, and compared with an inexpensive classical TF-IDF + Naïve Bayes setting. Several main findings emerge: (i) We find a specialization-generalization trade-off and asymmetric cross-register transfer: informal \(\rightarrow \) → formal degrades far less than formal \(\rightarrow \) → informal; (ii) misclassifications are structured: confusions concentrate in linguistically proximate pairs (e.g., Mexico-Colombia) and align with a low-dimensional geometry from Jensen-Shannon distances; (iii) errors stem from two main sources: over-reliance on toponyms/proper nouns (often defaulting to Spain), and interference from shared slang/morphology.