Learning compositional structures for semantic graph parsing

Jonas Groschwitz, Meaghan Fowlie, Alexander Koller


Abstract
AM dependency parsing is a method for neural semantic graph parsing that exploits the principle of compositionality. While AM dependency parsers have been shown to be fast and accurate across several graphbanks, they require explicit annotations of the compositional tree structures for training. In the past, these were obtained using complex graphbank-specific heuristics written by experts. Here we show how they can instead be trained directly on the graphs with a neural latent-variable model, drastically reducing the amount and complexity of manual heuristics. We demonstrate that our model picks up on several linguistic phenomena on its own and achieves comparable accuracy to supervised training, greatly facilitating the use of AM dependency parsing for new sembanks.
Anthology ID:
2021.spnlp-1.3
Volume:
Proceedings of the 5th Workshop on Structured Prediction for NLP (SPNLP 2021)
Month:
August
Year:
2021
Address:
Online
Venue:
spnlp
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
22–36
Language:
URL:
https://aclanthology.org/2021.spnlp-1.3
DOI:
10.18653/v1/2021.spnlp-1.3
Bibkey:
Cite (ACL):
Jonas Groschwitz, Meaghan Fowlie, and Alexander Koller. 2021. Learning compositional structures for semantic graph parsing. In Proceedings of the 5th Workshop on Structured Prediction for NLP (SPNLP 2021), pages 22–36, Online. Association for Computational Linguistics.
Cite (Informal):
Learning compositional structures for semantic graph parsing (Groschwitz et al., spnlp 2021)
Copy Citation:
PDF:
https://aclanthology.org/2021.spnlp-1.3.pdf
Video:
 https://aclanthology.org/2021.spnlp-1.3.mp4
Code
 coli-saar/am-parser