A Scalable Tool for Measuring Manner and Result Verbs in Developmental Language Research

In three linesComputational tool to classify manner and result verbs at scale. Uses linguistically informed prompts with LLMs to generate annotations over MASC and InterCorp data (436 VerbNet classes). RoBERTa-based classifier achieves 89.6% accuracy on three held-out gold-standard datasets. Applicable to developmental research on verb semantics.

Summary generated by Claude — human-verified