---
title: "Supervised Syntax-based Alignment between English Sentences"
description: "AMR corpora give sentence–graph pairs but do not record which English words produced which… 19 references, cited by 3 papers here."
authors: ["Chenhui Chu", "Sadao Kurohashi"]
year: 2016
venue: "arXiv:1606.02126"
source: "https://arxiv.org/abs/1606.02126"
doi: "10.48550/arxiv.1606.02126"
arxiv: "1606.02126"
licence: "arXiv — non-exclusive license to distribute"
licence_url: "http://arxiv.org/licenses/nonexclusive-distrib/1.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/chu-2016-supervised-syntax-based-alignment-between/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/chu-2016-supervised-syntax-based-alignment-between/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2016](https://md.research.elamur.ai/free/en/papers/2016/page/4/) / Supervised Syntax-based Alignment between English Sentences and Abstract Meaning Representation Graphs

# Supervised Syntax-based Alignment between English Sentences and Abstract Meaning Representation Graphs

**Authors:** [Chenhui Chu](https://md.research.elamur.ai/free/en/authors/chenhui-chu/) and [Sadao Kurohashi](https://md.research.elamur.ai/free/en/authors/sadao-kurohashi/) · **Year:** 2016 · **Venue:** arXiv:1606.02126 · **Licence:** [arXiv — non-exclusive license to distribute](http://arxiv.org/licenses/nonexclusive-distrib/1.0/)

**Links:** [Publisher's page](https://arxiv.org/abs/1606.02126) · [PDF](https://arxiv.org/pdf/1606.02126) · [DOI 10.48550/arxiv.1606.02126](https://doi.org/10.48550/arxiv.1606.02126) · [arXiv 1606.02126](https://arxiv.org/abs/1606.02126) · [OpenAlex](https://openalex.org/W2417260644)

**Citations:** 3 citations (OpenAlex, 2026-10-14) · [3](https://md.research.elamur.ai/free/en/papers/chu-2016-supervised-syntax-based-alignment-between/cited-by/) in our library, of the 141,840 papers whose bibliographies we read

## Summary
AMR corpora give sentence–graph pairs but do not record which English words produced which concepts and roles. Parsers need those links for training, so they must be inferred automatically. The earlier data-driven approach (ISI) flattened the AMR graph into a string and ran unsupervised IBM models on it. With only about 13k pairs available, that approach suffers badly from sparse statistics.

Chu and Kurohashi instead treat the task as aligning two trees:
- The English side is a Berkeley constituency parse.
- The AMR side is converted into a pseudo-constituency tree using the procedure of Pust et al.
- A supervised discriminative aligner (Nile, following Riesa et al.) then builds alignments bottom-up with a beam.
- Its features fire on syntactic categories as well as words, so evidence generalizes across tokens that rarely co-occur.

Trained on only 100 gold-aligned pairs, the aligner beats ISI by 1.7 F-score points. Feeding its links to the ISI AMR parser raises Smatch by 0.4. The work shows that a small amount of supervision combined with syntactic abstraction can outperform unsupervised alignment trained on much larger data.

## Key points
- The largest public English–AMR set at the time had 13,050 pairs. This is far smaller than typical MT word-alignment data.
- Alignments obey the constraint defined by Pourdamghani et al. (2014): each concept or role links to at most one English token, while a token may link to many.
- Both trees are preprocessed before alignment:
  - stop words and rarely aligned special concepts/roles are removed;
  - leaves are lowercased and stemmed to four letters.
- Alignment F-score improves by 1.7 points over the ISI aligner, using only 100 gold-aligned training pairs.
- AMR parsing on the authors' split improves from 64.7% to 65.1% Smatch with the Pust et al. parser. The paper cites human inter-annotator agreement at roughly 79–83 Smatch. JAMR (59.0%) and CAMR (61.3%) score lower on the same split.

## Interesting ideas
- Backing off from words to syntactic labels lets a pair never seen together still align, because their categories have been seen together; for example, the role ":topic" aligns to the preposition "with" through the label pair Ctopic/IN.
- Learned features can be negative as well as positive, so the model can block alignments that unsupervised co-occurrence wrongly favours; for example, "thing" is kept from attaching to "how" because their categories (Sstatement vs. WHNP) rarely co-occur.
- Converting a semantic graph into a tree-like shape (cutting extra parents, pushing role labels down to leaves) makes tree-to-tree machinery usable even though the two trees are not isomorphic.
- A beam of scored partial alignments is merged bottom-up through the syntax tree. This lets the model combine local features with non-local ones, including outputs of a third-party aligner.
- The error analysis exposes a concrete failure: a feature that favours aligning identical tokens linked the wrong "i", because the correct link had been removed in preprocessing.

## For Mova

How these ideas could serve Mova, our small open language model — its files, stages and tests:

- **I1** (frames): When mapping prepositions and clauses to VerbNet roles, score the pairing of UD relation/POS with role class, not only the specific word, so rare prepositions still get sensible roles.
- **I2** (learn): Include learned negative evidence when agreeing silver data across witness parsers, so frequent-but-wrong pairings are rejected rather than reinforced.
- **I3** (back): Linearize the meaning graph into a tree aligned with the UD tree, which would let generation and the text → graph → text compressor reuse tree-based alignment to check round-trips.
- **I4** (answer): Keep k-best partial interpretations per subtree when linking question phrases to world entities, so later global evidence can override an early local choice.
- **I5** (learn): Add negative controls that test whether a pipeline filtering step deletes the gold answer, so repair rules can be blamed on the right stage.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,086 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,889 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 9,296 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 9,206 works

🧭 Topics (3): [English](https://md.research.elamur.ai/free/en/topics/english/) · 1,373 works · [🧾 Semantic parsing & meaning representation](https://md.research.elamur.ai/free/en/topics/semantic-parsing/) · 135 works · [📐 Syntax & grammar formalisms](https://md.research.elamur.ai/free/en/topics/syntax-grammar/) · 161 works

## Cite

Cite as: Chenhui Chu and Sadao Kurohashi. 2016. Supervised Syntax-based Alignment between English Sentences and Abstract Meaning Representation Graphs. arXiv:1606.02126.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@misc{chu2016supervised,
  title = {{Supervised Syntax-based Alignment between English Sentences and Abstract Meaning Representation Graphs}},
  author = {Chenhui Chu and Sadao Kurohashi},
  year = {2016},
  doi = {10.48550/arxiv.1606.02126},
  eprint = {1606.02126},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1606.02126}
}
```

- [← previous Parallel Texts in the Hebrew Bible, New Methods and Visualizations](https://md.research.elamur.ai/free/en/papers/naaijer-2016-parallel-texts-hebrew-bible-new/)
- [next → Vocal Pathologies Detection and Mispronounced Phonemes Identification: Case of Arabic Continuous Speech](https://md.research.elamur.ai/free/en/papers/terbeh-2016-vocal-pathologies-detection-mispronounced-phonemes/)
- [↑ back to papers of 2016](https://md.research.elamur.ai/free/en/papers/2016/page/4/)

## References

19 references, 11 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. [Broad-coverage ccg semantic parsing with amr](https://md.research.elamur.ai/free/en/papers/artzi-2015-broad-coverage-ccg-semantic-parsing/). In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 1699–1710, Lisbon, Portugal, September 2015.
2. Laura Banarescu, Claire Bonial, Shu Cai, Madalina Georgescu, Kira Griffitt, Ulf Hermjakob, Kevin Knight, Philipp Koehn, Martha Palmer, and Nathan Schneider. [Abstract meaning representation for sembanking](https://md.research.elamur.ai/free/en/papers/banarescu-2013-abstract-meaning-representation-sembanking/). In Proceedings of the 7th Linguistic Annotation Workshop and Interoperability with Discourse, pages 178–186, Sofia, Bulgaria, August 2013.
3. Peter F. Brown, Stephen A. Della Pietra, Vincent J. Della Pietra, and Robert L. Mercer. The mathematics of statistical machine translation: Parameter estimation. Computational Linguistics, 19(2):263–312, 1993. [find](https://scholar.google.com/scholar?q=Peter+F%2E+Brown%2C+Stephen+A%2E+Della+Pietra%2C+Vincent+J%2E+Della+Pietra%2C+and+Robert+L%2E+Mercer%2E+The+mathematics+of+statistical+machine+translation%3A+Parameter+estimation%2E+Computational+Linguistics%2C+19%282%29%3A263%E2%80%933)
4. Shu Cai and Kevin Knight. Smatch: an evaluation metric for semantic feature structures. In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 748–752, Sofia, Bulgaria, August 2013. [find](https://scholar.google.com/scholar?q=Shu+Cai+and+Kevin+Knight%2E+Smatch%3A+an+evaluation+metric+for+semantic+feature+structures%2E+In+Proceedings+of+the+51st+Annual+Meeting+of+the+Association+for+Computational+Linguistics+%28Volume+2%3A+Short+Pape)
5. Michael Collins. [Discriminative training methods for hidden markov models: Theory and experiments with perceptron algorithms](https://md.research.elamur.ai/free/en/papers/collins-2002-discriminative-training-methods-hidden-markov/). In Proceedings of the 2002 Conference on Empirical Methods in Natural Language Processing, pages 1–8, July 2002.
6. Jeffrey Flanigan, Sam Thomson, Jaime Carbonell, Chris Dyer, and Noah A. Smith. [A discriminative graph-based parser for the abstract meaning representation](https://md.research.elamur.ai/free/en/papers/flanigan-2014-discriminative-graph-based-parser-abstract/). In Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1426–1436, Baltimore, Maryland, June 2014.
7. Daniel Gildea and Daniel Jurafsky. [Automatic labeling of semantic roles](https://md.research.elamur.ai/free/en/papers/gildea-2002-automatic-labeling-semantic-roles/). In Proceedings of the 38th Annual Meeting of the Association for Computational Linguistics, pages 512–520, Hong Kong, October 2000.
8. Dipendra Kumar Misra and Yoav Artzi. Neural shift-reduce ccg semantic parsing. In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 1775–1786, Austin, Texas, November 2016. [find](https://scholar.google.com/scholar?q=Dipendra+Kumar+Misra+and+Yoav+Artzi%2E+Neural+shift%2Dreduce+ccg+semantic+parsing%2E+In+Proceedings+of+the+2016+Conference+on+Empirical+Methods+in+Natural+Language+Processing%2C+pages+1775%E2%80%931786%2C+Austin%2C+Texas)
9. Franz Josef Och and Hermann Ney. A systematic comparison of various statistical alignment models. Computational Linguistics, 29(1):19–51, 2003. [↗](https://doi.org/10.1162/089120103321337421)
10. Martha Palmer, Daniel Gildea, and Paul Kingsbury. [The proposition bank: An annotated corpus of semantic roles](https://md.research.elamur.ai/free/en/papers/palmer-2005-proposition-bank-annotated-corpus-semantic/). Comput. Linguist., 31(1):71–106, March 2005. [↗](https://doi.org/10.1162/0891201053630264)
11. Xiaochang Peng, Chuan Wang, Daniel Gildea, and Nianwen Xue. Addressing the data sparsity issue in neural amr parsing. In Proceedings of the 15 th European Chapter of the Association for Computational Linguistics, Valencia, Spain, April 2017. [find](https://scholar.google.com/scholar?q=Xiaochang+Peng%2C+Chuan+Wang%2C+Daniel+Gildea%2C+and+Nianwen+Xue%2E+Addressing+the+data+sparsity+issue+in+neural+amr+parsing%2E+In+Proceedings+of+the+15+th+European+Chapter+of+the+Association+for+Computational+)
12. Slav Petrov and Dan Klein. [Improved inference for unlexicalized parsing](https://md.research.elamur.ai/free/en/papers/petrov-2007-improved-inference-unlexicalized-parsing/). In Human Language Technologies 2007: The Conference of the North American Chapter of the Association for Computational Linguistics; Proceedings of the Main Conference, pages 404–411, Rochester, New York, April 2007.
13. Nima Pourdamghani, Yang Gao, Ulf Hermjakob, and Kevin Knight. [Aligning english strings with abstract meaning representation graphs](https://md.research.elamur.ai/free/en/papers/pourdamghani-2014-aligning-english-strings-abstract-meaning/). In Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 425–429, Doha, Qatar, October 2014.
14. Michael Pust, Ulf Hermjakob, Kevin Knight, Daniel Marcu, and Jonathan May. [Parsing english into abstract meaning representation using syntax-based machine translation](https://md.research.elamur.ai/free/en/papers/pust-2015-parsing-english-into-abstract-meaning/). In Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 1143–1154, Lisbon, Portugal, September 2015.
15. Jason Riesa, Ann Irvine, and Daniel Marcu. Feature-rich language-independent syntax-based alignment for statistical machine translation. In Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing, pages 497–507, Edinburgh, Scotland, UK., July 2011. [find](https://scholar.google.com/scholar?q=Jason+Riesa%2C+Ann+Irvine%2C+and+Daniel+Marcu%2E+Feature%2Drich+language%2Dindependent+syntax%2Dbased+alignment+for+statistical+machine+translation%2E+In+Proceedings+of+the+2011+Conference+on+Empirical+Methods+in+N)
16. Chuan Wang, Nianwen Xue, and Sameer Pradhan. [Boosting transition-based amr parsing with refined actions and auxiliary analyzers](https://md.research.elamur.ai/free/en/papers/wang-2015-boosting-transition-based-amr-parsing/). In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 2: Short Papers), pages 857–862, Beijing, China, July 2015.
17. Keenon Werling, Gabor Angeli, and Christopher D. Manning. Robust subgraph generation improves abstract meaning representation parsing. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers), pages 982–991, Beijing, China, July 2015. [find](https://scholar.google.com/scholar?q=Keenon+Werling%2C+Gabor+Angeli%2C+and+Christopher+D%2E+Manning%2E+Robust+subgraph+generation+improves+abstract+meaning+representation+parsing%2E+In+Proceedings+of+the+53rd+Annual+Meeting+of+the+Association+for+)
18. Ying Zhang, Stephan Vogel, and Alex Waibel. Interpreting bleu/nist scores: How much improvement do we need to have a better system? In Proceedings of the Fourth International Conference on Language Resources and Evaluation (LREC-2004), Lisbon, Portugal, May 2004. ACL Anthology Identifier: L04-1489. [↗](https://aclanthology.org/L04-1489/)
19. Junsheng Zhou, Feiyu Xu, Hans Uszkoreit, Weiguang QU, Ran Li, and Yanhui Gu. [Amr parsing with an incremental joint model](https://md.research.elamur.ai/free/en/papers/zhou-2016-amr-parsing-incremental-joint-model/). In Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 680–689, Austin, Texas, November 2016.

*Updated 24 October 2026*