This is the repository for the Argument Structure Construction (ASC) Treebank. The ASC Treebank was used to train a Spacy transformer model that annotates ASCs with a high degree of accuracy (L1: F1 = .912; L2-written: F1 = .915; L2-spoken: F1 = .928).
Learned correspondences between form and function vary in terms of complexity and abstractionm, and so do the examples of the ASCs. There is currently no definitive set of ASCs that should be included in an ASC tagset. The current study drew on a range of previous literature and selected nine ASC types as a tagset.
ASC tag | Syntactic frame (example) | Semantic frame (example) |
---|---|---|
ATTR | Subj-Vcopular-NP | Xtheme is Yattribute |
INTRAN_S | Subj-V | Xagent acts |
INTRAN_MOT | Subj-V-Adverbial Particle/PP | Xtheme moves Ypath |
INTRAN_RES | Subj-V-NP/AdjP | Xpatient becomes Yresult |
TRAN_S | Subj-V-Obj | Xagent acts on Ypatient |
DITRAN | Subj-V-Objindirect-Objdirect | Xagent causes Yrecipient to receive Ztheme |
CAUS_MOT | Subj-V-Obj-PP | Xagent causes Ytheme to move Zpath/goal |
TRAN_RES | Subj-V-Obj-NP | Xagent causes Ytheme to become Zstate |
PASSIVE | Subj-auxVpast participle(-by-PP) | Xtheme undergo V (by Yagent) |
- The documenation for the ASC-Treebank (v1) is available online.
- The more detailed version of the documentation (v2) is available in PDF format.
- The annotation model (trained on the ASC treebank v1) (including a demo) is available here.
- The ASC tagger demo (trained on the ASC treebank v2) is available here (updated Oct. 2024).
If you use our treebanks or models, please refer to the following papers:
- (ASC treebank + model v1,) 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.
- (ASC treebank v2) Sung, H., & Kyle, K. (2024). Annotation Scheme for English Argument Structure Constructions Treebank. In Proceedings of The 18th Linguistic Annotation Workshop (LAW-XVIII), 12-18, Association for Computational Linguistics.
- (ASC model v2) Sung, H., & Kyle, K. (in press). Leveraging pre-trained language models for linguistic analysis: A case of argument structure constructions. In the Association for Computational Linguistics: EMNLP 2024.
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.