Generation from Abstract Meaning Representation using Tree Transducers

What it is about

The paper generates English from AMR graphs in two steps. A breadth-first spanning tree is taken from the graph, and a weighted tree-to-string transducer turns that tree into a sentence. The transducer rules are extracted from AMR-aligned sentences: each aligned subgraph becomes a rule with argument slots for its children, and synthetic rules generalise over concept and argument realisations. Decoding is a linear model over rule features plus a 5-gram language model, tuned with MERT. On AMR 1.0 (about 10,000 training sentences) the full system reaches 22.1 BLEU on test. Removing the synthetic rules drops it to 9.1; removing the basic or abstract rules changes almost nothing.

Ideas

• I1. Extract a basic rule for each meta token from the span of source words it was read from, with argument slots for its children. This is the same contract as the encoder's context table in profile.rs. Test whether the table covers unseen (concept, relation) pairs as well as the paper's synthetic model does.
• I2. Run the translator with each rule source removed in turn (standard table, context table, bare-form fallback) and report the bit cost per class. Each component should earn its place by a measured residual on dev.
• I3. The spanning tree is a fixed breadth-first order with sorted child labels, which the authors call a baseline. Test whether the child order chosen by the tree reader changes the residual; run twice to check identical output.
• I4. The argument rule for a relation such as destination or purpose fixes a preposition. Count (concept, relation) against function words over the Bible pairs in a BTreeMap, and use the result as the function-word feature of open problem 4.

For Mova

The ideas touch tools/translate/profile.rs (encoder), en/src/lin.rs (word order) and open problem 4 (function words). The core of the method is a rule table with fallback, which carries over without change. The paper's MERT tuning becomes counts with thresholds, checked on dev.

Topics: Machine translation

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.18653/v1/n16-1087

@inproceedings{flanigan2016generation,
  title = {{Generation from Abstract Meaning Representation using Tree Transducers}},
  author = {Flanigan, Jeffrey and Dyer, Chris and Smith, Noah A. and Carbonell, Jaime},
  booktitle = {Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies},
  year = {2016},
  doi = {10.18653/v1/n16-1087},
  url = {https://doi.org/10.18653/v1/n16-1087}
}

References

25 references, 11 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. 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
2. Anja Belz, Michael White, Dominic Espinosa, Eric Kow, Deirdre Hogan, and Amanda Stent. 2011. The first surface realisation shared task: Overview and evaluation results. In Proc. of the 13th European Workshop on Natural Language Generation. find our card
3. Bernd Bohnet, Leo Wanner, Simon Mille, and Alicia Burga. 2010. Broad coverage multilingual deep sentence generation with a stochastic multi-level realizer. In Proc. of COLING. find our card
4. Aoife Cahill and Josef Van Genabith. 2006. Robust pcfg-based generation using automatically acquired LFG approximations. In Proc. of COLING-ACL. find
5. Michael Collins. 2002. Discriminative training methods for hidden Markov models: Theory and experiments with perceptron algorithms. In Proc. of EMNLP. find our card
6. John Duchi, Elad Hazan, and Yoram Singer. 2011. Adaptive subgradient methods for online learning and stochastic optimization. JMLR, 12. find
7. Chris Dyer, Adam Lopez, Juri Ganitkevitch, Johnathan Weese, Ferhan Ture, Phil Blunsom, Hendra Setiawan, Vladimir Eidelman, and Philip Resnik. 2010. cdec: A decoder, alignment, and learning framework for finite-state and context-free translation models. In Proc. of ACL. find
8. 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
9. Michel Galley, Mark Hopkins, Kevin Knight, and Daniel Marcu. 2004. What’s in a translation rule? In Proc. of HLT-NAACL. find our card
10. Joshua Goodman. 1999. Semiring parsing. CL, 25(4). find
11. Jonathan Graehl and Kevin Knight. 2004. Training tree transducers. In Proc. of HLT-NAACL. find our card
12. Kenneth Heafield, Ivan Pouzyrevsky, Jonathan H. Clark, and Philipp Koehn. 2013. Scalable modified Kneser-Ney language model estimation. In Proc. of ACL. find
13. Deirdre Hogan, Conor Cafferkey, Aoife Cahill, and Josef Van Genabith. 2007. Exploiting multi-word units in history-based probabilistic generation. In Proc. of ACL. find
14. Liang Huang, Kevin Knight, and Aravind Joshi. 2006. Statistical syntax-directed translation with extended domain of locality. In Proc. of AMTA. find our card
15. Bevan Jones, Jacob Andreas, Daniel Bauer, Karl Moritz Hermann, and Kevin Knight. 2012. Semantics-based machine translation with hyperedge replacement grammars. In Proc. of COLING. find our card
16. Philipp Koehn, Hieu Hoang, Alexandra Birch, Chris Callison-Burch, Marcello Federico, Nicola Bertoldi, Brooke Cowan, Wade Shen, Christine Moran, Richard Zens, et al. 2007. Moses: Open source toolkit for statistical machine translation. In Proc. of ACL. find
17. Irene Langkilde and Kevin Knight. 1998. Generation that exploits corpus-based statistical knowledge. In Proc. of COLING-ACL. find
18. Irene Langkilde. 2000. Forest-based statistical sentence generation. In Proc. of NAACL 2000. find
19. Hiroko Nakanishi, Yusuke Miyao, and Jun’ichi Tsujii. 2005. Probabilistic models for disambiguation of an HPSG-based chart generator. In Proc. of IWPT. find
20. Franz Josef Och. 2003. Minimum error rate training in statistical machine translation. In Proc. of ACL. find
21. Martha Palmer, Daniel Gildea, and Paul Kingsbury. 2005. The proposition bank: An annotated corpus of semantic roles. CL, 31(1). find our card
22. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proc. of ACL. find our card
23. Frank Rosenblatt. 1957. The perceptron—a perceiving and recognizing automaton. Technical Report 85-460-1, Cornell Aeronautical Laboratory. find
24. Michael White, Rajakrishnan Rajkumar, and Scott Martin. 2007. Towards broad coverage surface realization with CCG. In Proc. of the Workshop on Using Corpora for NLG: Language Generation and Machine Translation. find our card
25. Yuk Wah Wong and Raymond J Mooney. 2006. Learning for semantic parsing with statistical machine translation. In Proc. of HLT-NAACL. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/flanigan-2016-generation-abstract-meaning-representation-tree/