Główne pojęcia
Tools for automatic annotation offer an effective alternative to manual annotation, aiding in large-scale studies on child language acquisition.
Streszczenie
Introduction
The central question of grammar acquisition.
Traditional reliance on manual annotations.
Proposal for automatic coding scheme.
Contributions
New coding scheme for grammaticality.
Training and evaluation of NLP models.
Related Work
Supervised approaches for grammaticality annotation.
Manual Annotation
Development of annotation scheme.
Data
Transcripts from English CHILDES corpus annotated.
Automatic Annotation
Range of models evaluated, with Transformer-based models performing best.
Results
Evaluation metrics show performance compared to human annotators.
Analyses
Effect of context length and training data size explored.
Limitations
Challenges in fine-grained error analysis and dialect variations discussed.
Statystyki
We annotate more than 4,000 utterances from a large corpus of transcribed conversations.
Our results show that fine-tuned Transformer-based models perform best.