ASC-Treebank Documentation

In cognitive linguistics, a construction represents a form-meaning pair. Constructions can exist at the syntactic/lexicogrammatical level, wherein a verb and its argument structure constitute a form that corresponds to a propositional meaning (e.g., Diessel, 2004; Fillmore, Kay, & O’Connor, 1988; Goldberg, 1995, 2003, 2006; Jackendoff, 2002). These constructions are referred to as argument structure constructions (ASCs). The current project represents the construction of a gold-standard annotated treebank of ASCs, which was used to train a Spacy transformer model that annotates ASCs with a high degree of accuracy (F1 = .918). The annotation model (including a demo) is available here.

✅ The annotation model (including a demo) is currently available here.

Citation

If you use the model or the treebank, please cite this paper:

✅ Kyle, K. & Sung, H. (2023). An argument structure construction treebank, In Proceedings of the First International Workshop on Construction Grammars and NLP (CxGs+NLP, GURT/SyntaxFest 2023), 51–62, Association for Computational Linguistics.

Contents

Introduction

• Quick overview
  • ASC Tagset
  • Dataset
  • Creating an ASC treebank
  • ASC representation in Treebank
• ASC guideline
  • ATTR
  • INTRAN_S
  • INTRAN_MOT
  • INTRAN_RES
  • TRAN_S
  • DITRAN
  • CAUS_MOT
  • TRAN_RES
  • PASSIVE
  • * Notes
• Model performance
  • Model training
  • Transformer model
  • F1 Scores for all models

Resources

• Related projects

License

This documentation and its content are licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.