New theory suggests universal syntax prior emerges from language production

A new paper proposes a universal prior over syntactic structures that emerges from a model of incremental language production, rather than solely from language-specific data. This prior, represented as dependency trees, assigns probabilities to syntactic structures without fitting parameters to linguistic corpora. The model demonstrates that this prior assigns higher probabilities to attested trees than random ones across 138 diverse languages and correlates positively with corpus-estimated probabilities in 33 out of 34 languages. The findings suggest that syntactic structure probabilities may be partially shaped by language production processes, providing a data-independent bias for probabilistic language models. AI

IMPACT Suggests a cognitive origin for syntactic structure probabilities, potentially informing future probabilistic language models.

RANK_REASON Academic paper on a theoretical model of language syntax. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• CORE Recommender
• DagsHub
• Fermín Moscoso Del Prado Martín
• Gotit.pub
• Hugging Face
• Influence Flower
• ScienceCast

• paper
• other

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