Corpus choice significantly impacts language dependency estimates, study finds

A new study published on arXiv investigates how the choice of corpus affects dependency-distance estimates in natural language processing. Researchers compared estimates across 38 language treebanks from Universal Dependencies v2.18, finding that corpus substitution significantly altered results, reversing nearly 40 percent of pairwise language orderings. The study suggests that dependency-distance is more of a corpus-conditioned factor influenced by grammar, register, and annotation rather than a stable language-level parameter, though the general principle of dependency-length minimization was consistently observed. AI

IMPACT Highlights the sensitivity of linguistic analysis to data sources, impacting NLP model training and evaluation.

RANK_REASON The cluster contains a research paper published on arXiv detailing linguistic analysis methods. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• alphaXiv
• arXivLabs
• CatalyzeX Code Finder for Papers
• CORE Recommender
• DagsHub
• Data Language Models
• Gotit.pub
• Hugging Face
• McDonnell Douglas
• ScienceCast
• Universal Dependencies
• v2.18

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →