Broad Coverage Multilingual Deep Sentence Generation with a Stochastic Multi-Level Realizer

What it is about

The paper describes a multilingual generator that starts from deep, predicate-argument-like structures and produces sentences in several languages. It works in stages, and each stage has its own statistical decoder built from support vector machines. One decoder maps the semantic level to the syntactic level, one orders the words, and one produces the morphology. The stages are trained on multi-level annotated corpora. The authors report a BLEU score in the range of roughly 0.61 to 0.69 across the languages, with the highest value for Chinese, and they compare it with a grammar-based realiser on related inputs. They also observe that current annotation schemes do not always serve generation well, and that the corpora need extra generation-oriented layers.

Ideas

• I1. Score errors per stage. Keep the three stage splits (structure, order, morphology) as separate residual counts. The translator already reports residuals per layer and class, so this adds a finer split inside the encoder.
• I2. Use a stage-wise count table where the paper uses a classifier. Each stage decides from counts of its own features, which keeps every step inspectable and deterministic. Keep the table order in BTreeMap.
• I3. Check the annotation schemes against the generation need. Count how many meta tokens of the translator lack a head or relation that the encoder needs, as the paper's observation suggests.

For Mova

Only the principle carries over: a deep generator in stages, each stage scored on its own. The fit is tools/translate (profile.rs, the residual split) and en/src/lin.rs. The paper's SVM method is not for Mova, and a count-based form of the same stages is what Mova can use.

Topics: Machine translation

Cite

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

@inproceedings{bohnet2010broad,
  title = {{Broad Coverage Multilingual Deep Sentence Generation with a Stochastic Multi-Level Realizer}},
  author = {Bohnet, Bernd and Wanner, Leo and Mille, Simon and Burga, Alicia},
  booktitle = {Proceedings of the 23rd International Conference on Computational Linguistics (Coling 2010)},
  year = {2010},
  url = {https://aclanthology.org/C10-1012/}
}

References

25 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Bangalore, S. and O. Rambow. 2000. Exploiting a Probabilistic Hierarchical Model for Generation. In Proceedings of COLING ’00, pages 42–48. find
2. Bangalore, S., J. Chen, and O. Rambow. 2001. Impact of Quality and Quantity of Corpora on Stochastic Generation. In Proceedings of the EMNLP Conference, pages 159–166. find
3. Bateman, J.A., I. Kruijff-Korbayová, and G.-J. Kruijff. 2005. Multilingual Resource Sharing Across Both Related and Unrelated Languages: An Implemented, Open-Source Framework for Practical Natural Language Generation. Research on Language and Computation, 15:1–29. find
4. Belz, A. 2008. Automatic generation of weather forecast texts using comprehensive probabilistic generation-space models. Natural Language Engineering, 14(4):431–455. find
5. Bohnet, B. 2004. A graph grammar approach to map between dependency trees and topological models. In Proceedings of the IJCNLP, pages 636–645. find
6. Bröker, N. 1998. Separating Surface Order and Syntactic Relations in a Dependency Grammar. In Proceedings of the COLING/ACL ’98. find
7. Crammer, K., O. Dekel, S. Shalev-Shwartz, and Y. Singer. 2006. Online Passive-Aggressive Algorithms. Journal of Machine Learning Research, 7:551–585. find
8. Duchier, D. and R. Debusmann. 2001. Topological dependency trees: A constraint-based account of linear precedence. In Proceedings of the ACL. find
9. Elhadad, M. and J. Robin. 1996. An overview of SURGE: A reusable comprehensive syntactic realization component. Technical Report TR 96-03, Department of Mathematics and Computer Science, Ben Gurion University. find our card
10. Filippova, K. and M. Strube. 2008. Sentence fusion via dependency graph compression. In Proceedings of the EMNLP Conference. find
11. Filippova, K. and M. Strube. 2009. Tree linearization in English: Improving language model based approaches. In Proceedings of the NAACL ’09 and HLT, Short Papers, pages 225–228. find our card
12. Gerdes, K. and S. Kahane. 2001. Word order in German: A formal dependency grammar using a topological hierarchy. In Proceedings of the ACL. find
13. Hajič, J. et al. 2009. The CoNLL-2009 Shared Task: Syntactic and Semantic Dependencies in Multiple Languages. In Proceedings of the CoNLL. find our card
14. He, W., H. Wang, Y. Guo, and T. Liu. 2009. Dependency based chinese sentence realization. In Proceedings of the ACL and of the IJCNLP of the AFNLP, pages 809–816. find
15. Knight, K. and V. Hatzivassiloglou. 1995. Two-level, many paths generation. In Proceedings of the ACL. find
16. Langkilde, I. and K. Knight. 1998. Generation that exploits corpus-based statistical knowledge. In Proceedings of the COLING/ACL, pages 704–710. find
17. Langkilde-Geary, I. 2002. An empirical verification of coverage and correctness for a general-purpose sentence generator. In Proceedings of the Second INLG Conference, pages 17–28. find
18. Lavoie, B. and O. Rambow. 1997. A fast and portable realizer for text generation systems. In Proceedings of the 5th Conference on ANLP. find our card
19. Levenshtein, V.I. 1966. Binary codes capable of correcting deletions, insertions, and reversals. Soviet Physics, 10:707–710. find
20. Mellish, C., D. Scott, L. Cahill, D. Paiva, R. Evans, and M. Reape. 2006. A reference architecture for natural language generation systems. Natural Language Engineering, 12(1):1–34. find
21. Minnen, G., J. Carroll, and D. Pearce. 2001. Applied morphological processing for English. Natural Language Engineering, 7(3):207–223. find
22. Oh, A.H. and A.I. Rudnicky. 2000. Stochastic language generation for spoken dialogue systems. In Proceedings of the ANL/NAACL Workshop on Conversational Systems, pages 27–32. find
23. Palmer, M., D. Gildea, and P. Kingsbury. 2005. The proposition bank: An annotated corpus of semantic roles. Computational Linguistics, 31(1):71–105. find our card
24. Ringger, E., M. Gamon, R.C. Moore, D. Rojas, M. Smets, and S. Corston-Oliver. 2004. Linguistically informed statistical models of constituent structure for ordering in sentence realization. In Proceedings of COLING, pages 673–679. find
25. Wan, S., M. Dras, Dale R., and C. Paris. 2009. Improving Grammaticality in Statistical Sentence Generation: Introducing a Dependency Spanning Tree Algorithm with an Argument Satisfaction Model. In Proceedings of the EACL ’09, pages 852–860. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/bohnet-2010-broad-coverage-multilingual-deep-sentence/