A Structured Syntax-Semantics Interface for English-AMR Alignment

What it is about

The paper tests a common assumption, that AMR closely mirrors dependency syntax. The authors design an alignment framework from AMR graphs to enhanced Universal Dependencies graphs. Each AMR node or edge can be evoked by words, by syntactic structure, or by both. They hand-annotate 200 AMRs in this framework. Their annotation explains 97% of AMR edges. The earlier ISI scheme, which maps edges only to words, explains about 23%. Many edges are evoked by syntax rather than by a word: a role is licensed by a subject, not by a particular token. The annotators also find many places where AMR and syntax diverge. A baseline aligner has two passes: a lexical pass based on string similarity and negation rules, then a structural pass. It does worse on this task than on string alignment. The corpus is also used to find errors in an AMR parser, whose Smatch on these sentences is 0.65.

Ideas

• I1. Give each meta token two kinds of evidence: the words that evoke it, and the tree node or relation that licenses it. The reader already keeps source words, so the new part is a pointer to the licensing node. The gap between 97% and 23% of edges is the case for keeping both.
• I2. Hand-label a small set of verses (50 to 200) with the same three labels: word-evoked, syntax-evoked, or unexplained. The share of unexplained edges becomes a gate on the reader, measured on that set, and it does not depend on the English parser.
• I3. Write the paper's lexical pass as a deterministic rule list: a Levenshtein distance of at most 15% of the longer word, morphological negation, and the negation rule. These rules live in Rust, are counted in the report, and run before any learned lexicon.
• I4. Use the paper's error classes (faulty morphology, duplicated words, accidental string similarity, missed similarity) as the reader's residual classes, so that the reader's errors and its residual report use the same vocabulary.

For Mova

The ideas apply to tools/translate/meta.rs (the lexical pass and the source-word field of each meta token), tools/roundtrip (the tree that the reader uses), and tools/hbo and tools/macula (the UD trees of Hebrew, where the same two-kind evidence would apply). The paper's aligner is a simple baseline and its statistics are from 200 sentences. For Mova, the counts from the translation pairs are the larger source.

Topics: Machine translation

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.18653/v1/n18-1106

@inproceedings{szubert2018structured,
  title = {{A Structured Syntax-Semantics Interface for English-AMR Alignment}},
  author = {Szubert, Ida and Lopez, Adam and Schneider, Nathan},
  booktitle = {Proceedings of the 2018 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long Papers)},
  year = {2018},
  doi = {10.18653/v1/n18-1106},
  url = {https://doi.org/10.18653/v1/n18-1106}
}

References

32 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Yoav Artzi, Kenton Lee, and Luke Zettlemoyer. 2015. Broad-coverage CCG semantic parsing with AMR. In Proc. of EMNLP. find
2. Laura Banarescu, Claire Bonial, Shu Cai, Madalina Georgescu, Kira Griffitt, Ulf Hermjakob, Kevin Knight, Philipp Koehn, Martha Palmer, and Nathan Schneider. 2013. Abstract Meaning Representation for sembanking. In Proc. of the 7th Linguistic Annotation Workshop and Interoperability with Discourse. find our card
3. Emily M. Bender, Dan Flickinger, Stephan Oepen, Woodley Packard, and Ann A. Copestake. 2015. Layers of interpretation: On grammar and compositionality. In Proc. of IWCS. find
4. Roger Brown. 1973. A first language: The early stages. Harvard University Press. find
5. Shu Cai and Kevin Knight. 2013. Smatch: an evaluation metric for semantic feature structures. In Proc. of ACL. find
6. Danqi Chen and Christopher D. Manning. 2014. A fast and accurate dependency parser using neural networks. In Proc. of EMNLP. find our card
7. Wei-Te Chen and Martha Palmer. 2017. Unsupervised AMR-dependency parse alignment. In Proc. of EACL. find our card
8. Chenhui Chu and Sadao Kurohashi. 2016. Supervised syntax-based alignment between English sentences and Abstract Meaning Representation graphs. arXiv preprint http://arxiv.org/abs/1606.02126 . ↗
9. Marco Damonte, Shay B. Cohen, and Giorgio Satta. 2017. An incremental parser for Abstract Meaning Representation. In Proc. of EACL. find
10. Jeffrey Flanigan, Chris Dyer, Noah A. Smith, and Jaime G. Carbonell. 2016. Generation from Abstract Meaning Representation using tree transducers. In Proc. of HLT-NAACL. find our card
11. Jeffrey Flanigan, Sam Thomson, Jaime Carbonell, Chris Dyer, and Noah A. Smith. 2014. A discriminative graph-based parser for the Abstract Meaning Representation. In Proc. of ACL. find
12. Stig-Arne Grönroos, Sami Virpioja, Peter Smit, and Mikko Kurimo. 2014. Morfessor FlatCat: An HMM-based method for unsupervised and semi-supervised learning of morphology. In Proc. of COLING. find
13. Daniel Hershcovich, Omri Abend, and Ari Rappoport. 2017. A transition-based directed acyclic graph parser for UCCA. In Proc. of ACL. find
14. Kevin Knight, Laura Baranescu, Claire Bonial, Madalina Georgescu, Kira Griffitt, Ulf Hermjakob, Daniel Marcu, Martha Palmer, and Nathan Schneider. 2014. Abstract Meaning Representation (AMR) Annotation Release 1.0. Technical Report LDC2014T12, Linguistic Data Consortium, Philadelphia, Pennsylvania, USA. https:// catalog.ldc.upenn.edu/LDC2014T12 . ↗
15. Ioannis Konstas, Srinivasan Iyer, Mark Yatskar, Yejin Choi, and Luke Zettlemoyer. 2017. Neural AMR: Sequence-to-sequence models for parsing and generation. In Proc. of ACL. find our card
16. Geert-Jan M. Kruijff. 2006. Dependency grammar. In Keith Brown, editor, Encyclopedia of Language and Linguistics (Second Edition), Elsevier, Oxford, pages 444–450. find
17. Christopher D. Manning, Mihai Surdeanu, John Bauer, Jenny Finkel, Steven J. Bethard, and David Mc-Closky. 2014. The Stanford CoreNLP natural language processing toolkit. In Proc. of ACL System Demonstrations. find
18. David Mareček, Martin Popel, Loganathan Ramasamy, Jan Štěpánek, Daniel Zeman, Zdeněk Žabokrtský, and Jan Hajič. 2013. Cross-language study on influence of coordination style on dependency parsing performance. Technical Report 49, ÚFAL MFF UK. find
19. Dipendra Kumar Misra and Yoav Artzi. 2016. Neural shift-reduce CCG semantic parsing. In Proc. of EMNLP. find
20. Joakim Nivre. 2005. Dependency grammar and dependency parsing. Technical Report MSI report 05133, Växjö University School of Mathematics and Systems Engineering, Växjö, Sweden. find
21. Stephan Oepen, Marco Kuhlmann, Yusuke Miyao, Daniel Zeman, Dan Flickinger, Jan Hajič, Angelina Ivanova, and Yi Zhang. 2014. SemEval 2014 Task 8: Broad-Coverage Semantic Dependency Parsing. In Proc. of SemEval. find
22. Barbara H. Partee. 2014. A brief history of the syntax-semantics interface in Western formal linguistics. Semantics-Syntax Interface 1:1–20. find
23. Xiaochang Peng, Chuan Wang, Daniel Gildea, and Nianwen Xue. 2017. Addressing the data sparsity issue in neural AMR parsing. In Proc. of EACL. find
24. Nima Pourdamghani, Yang Gao, Ulf Hermjakob, and Kevin Knight. 2014. Aligning English strings with Abstract Meaning Representation graphs. In Proc. of EMNLP. find our card
25. Nima Pourdamghani, Kevin Knight, and Ulf Hermjakob. 2016. Generating English from Abstract Meaning Representations. In Proc. of INLG. find our card
26. Siva Reddy, Oscar Täckström, Michael Collins, Tom Kwiatkowski, Dipanjan Das, Mark Steedman, and Mirella Lapata. 2016. Transforming dependency structures to logical forms for semantic parsing. Transactions of the Association for Computational Linguistics 4:127–140. find
27. Sebastian Schuster and Christopher D. Manning. 2016. Enhanced English Universal Dependencies: an improved representation for natural language understanding tasks. In Proc. of LREC. find our card
28. Gabriel Stanovsky, Jessica Ficler, Ido Dagan, and Yoav Goldberg. 2016. Getting more out of syntax with PropS. arXiv preprint http://arxiv.org/ abs/1603.01648 . ↗
29. Rik van Noord and Johan Bos. 2017. Neural semantic parsing by character-based translation: experiments with Abstract Meaning Representations. arXiv preprint http://arxiv.org/abs/1705.09980 . ↗
30. Chuan Wang, Nianwen Xue, Sameer Pradhan, and Sameer Pradhan. 2015. A transition-based algorithm for AMR parsing. In Proc. of NAACL-HLT. find
31. Aaron Steven White, Drew Reisinger, Keisuke Sakaguchi, Tim Vieira, Sheng Zhang, Rachel Rudinger, Kyle Rawlins, and Benjamin Van Durme. 2016. Universal Decompositional Semantics on Universal Dependencies. In Proc. of EMNLP. find our card
32. Sheng Zhang, Rachel Rudinger, and Benjamin Van Durme. 2017. An evaluation of PredPatt and Open IE via stage 1 semantic role labeling. In Proc. of IWCS. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/szubert-2018-structured-syntax-semantics-interface-english/