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

In three linesComputational tool for classifying manner and result verbs at scale. Uses linguistically informed prompts with LLMs to generate annotations over MASC and InterCorp (436 VerbNet classes), then trains a RoBERTa-based classifier. Performance: 89.6% accuracy across three gold-standard datasets.

Summary generated by Claude — human-verified