Discourse Representation Structures (DRSs) are a meaning representation formalism based on Discourse Representation Theory (Kamp and Reyle, 1993). They are in some respects similar to Abstract Meaning Representations (AMR, Banarescu et al., 2013), but different in others, e.g., they explicitly represent accessibility and operator scope. Thanks to the Parallel Meaning Bank (Abzianidze et al., 2017), a sizeable corpus of sentences consistently annotated with DRSs is now available, and a shared task on semantic parsing with DRSs as the target representation is being organized at this year's IWCS (Abzianidze et al, 2019).
In this talk, I present my submission to this shared task, using a transition-based semantic parsing architecture that is in principle also applicable to other deep meaning representation formalisms, such as AMR and frame-based semantics à la Kallmeyer et al. (2015). On the spectrum that ranges from grammar-based semantic parsers (Copestake, 2002, Bos, 2008) to encoder/decoder-based semantic parsers (van Noord and Bos, 2017, Liu et al., 2018, van Noord et al. 2018), the transition-based approach aims to occupy a useful middle ground: on one hand, it uses an explicit lexicon of word-meaning pairs, amenable to tweaking by linguists and engineers, and to interfacing with rule-based components. On the other hand, it eschews grammar engineering and uses a neural network to learn to assemble word meanings into sentence meanings. I surmise that this is an architecture especially useful in production, semi-automatic annotation, and education settings.
The approach is inspired by the AMR parser of Ballesteros and Al-Onaizan (2017) and, by extension, the non-projective dependency parsing algorithm of Nivre (2009): it uses a transition system to process tokens from left to right, and stack-LSTMs to create vector representations of parser states to make transition decisions. To apply this approach to DRSs, atomic node labels are replaced by lexical clause lists (LCLs) and edge labels by sets of referent address pairs (RAPs), which encode decisions to unify specific discourse referents. The lexicon is factored to deal with sparse data, and various preprocessing and postprocessing steps are applied to ease learning.
In this talk, I present the architecture and first results, and discuss possible future application to other meaning representation formalisms.
Abzianidze, L., J. Bjerva, K. Evang, H. Haagsma, R. van Noord, P. Ludmann, D.-D. Nguyen, and J. Bos (2017). The parallel meaning bank: Towards a multilingual corpus of translations annotated with compositional meaning representations. In Proceedings of the 15th Conference of the European Chapter of the Association for Computational Linguistics: Volume 2, Short Papers, pp. 242–247. Association for Computational Linguistics.
Ballesteros, M. and Y. Al-Onaizan (2017). AMR parsing using stack-LSTMs. In Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pp. 1269–1275. Association for Computational Linguistics.
Bos, J. (2008). Wide-coverage semantic analysis with Boxer. In Semantics in Text Processing. STEP 2008 Conference Proceedings.
Kallmeyer, L., Lichte, T., Osswald, R., Pogodalla, S., & Wurm, C. (2015, August). Quantification in frame semantics with hybrid logic. In Proceedings of the Type Theory and Lexical Semantics (TYTLES) ESSLLI workshop.
Liu, J., S. B. Cohen, and M. Lapata (2018). Discourse representation structure parsing. In Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 429–439. Association for Computational Linguistics.
Nivre, J. (2009). Non-projective dependency parsing in expected linear time. In Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International Joint Conference on Natural Language Processing of the AFNLP, pp. 351–359. Association for Computational Linguistics.
van Noord, R., L. Abzianidze, A. Toral, and J. Bos (2018). Exploring neural methods for parsing discourse representation structures. Transactions of the Association for Computational Linguistics 6, 619–633.
van Noord, R. and J. Bos (2017). Neural semantic parsing by character-based translation: Experiments with abstract meaning representations. Computational Linguistics in the Netherlands Journal 7, 93–108.