Integrating shallow-transfer rules into phrase-based statistical machine translation

What it is about

A hybrid method that feeds a phrase-based statistical system with the knowledge of a shallow-transfer rule system. Each Apertium dictionary entry, and each span of text matched by a structural transfer rule, is turned into a bilingual phrase pair, and the pairs are added to the phrase table. Three ways of scoring the new pairs are compared: appending them to the corpus, adding them once to the table with a boolean flag, or both. Experiments use Spanish-English in both directions on Europarl training sizes from 5,000 sentences up to all of it, in-domain and out-of-domain. The gain is largest when training data is small and the test text is general news that the rules cover. For Spanish-English out-of-domain with 5,000 training sentences, BLEU rises from 14.33 for the baseline to 21.32 with phrase-rules; with the whole corpus the gain is smaller (24.23 to 24.81). Generating only the spans that occur in the input avoids meaningless combinations.

Ideas

• I1. Add a boolean feature to the translator's encoder table that marks chunks coming from curated author deltas, as the phrase-rules variant does. A curated chunk should not be drowned by counts of the corpus, so the scorer needs to know where a chunk came from.
• I2. Generate candidate multi-word chunks only from spans that occur in the input text, not from every combination of dictionary entries. For the encoder, this means building chunks from the anchors of the frames in the sentence, which avoids the combinatorial junk the paper warns about.
• I3. Report gains by training size and by domain, as the paper does. For the translator, show the curve of exact-match against the number of verse pairs, separately for verses that are narrative and verses that are poetic or legal.
• I4. Give multi-word units a fixed alignment instead of letting the aligner split them. The paper leaves them unaligned and gives them a weight of 1.0. In the Hebrew reader, a dictionary multi-word unit could be fixed before IBM Model 1 runs.

For Mova

The code is tools/translate (ibm.rs for alignment, profile.rs for the encoder table, hebrew.rs for the reader). The phrase-pair generation idea fits the encoder; the boolean-flag scoring is cheap to test in profile.rs. Mova is not a phrase-based system, so none of the SMT scoring is copied, and the claims above must be measured on the Bible pairs before they are stated.

Topics: Machine translation

Cite

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

@inproceedings{sanchezcartagena2011integrating,
  title = {{Integrating shallow-transfer rules into phrase-based statistical machine translation}},
  author = {Sánchez-Cartagena, Víctor M. and Sánchez-Martínez, Felipe and Pérez-Ortiz, Juan Antonio},
  booktitle = {Proceedings of Machine Translation Summit XIII: Papers},
  year = {2011},
  url = {https://aclanthology.org/2011.mtsummit-papers.64/}
}

References

21 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. P. F. Brown, S. A. D. Pietra, V. J. D. Pietra, M. J. Goldsmith, J. Hajic, R. L. Mercer, and S. Mohanty. 1993. But dictionaries are data too. In Proceedings of the workshop on Human Language Technology, pages 202–205. find
2. C. Callison-Burch, P. Koehn, C. Monz, and O. Zaidan. 2011. Findings of the 2011 workshop on statistical machine translation. In Proceedings of the Sixth Workshop on Statistical Machine Translation, pages 22–64. find
3. L. Dugast, J. Senellart, and P. Koehn. 2007. Statistical post-editing on SYSTRAN’s rule-based translation system. In Proceedings of the Second Workshop on Statistical Machine Translation, pages 220–223. find
4. L. Dugast, J. Senellart, and P. Koehn. 2008. Can we Relearn an RBMT System? In Proceedings of the Third Workshop on Statistical Machine Translation, pages 175–178. find
5. A. Eisele, C. Federmann, H. Saint-Amand, M. Jellinghaus, T. Herrmann, and Y. Chen. 2008. Using Moses to integrate multiple rule-based machine translation engines into a hybrid system. In Proceedings of the Third Workshop on Statistical Machine Translation, pages 179–182. find
6. M.L. Forcada, M. Ginestı́-Rosell, J. Nordfalk, J. O’Regan, S. Ortiz-Rojas, J.A. Pérez-Ortiz, F. Sánchez-Martı́nez, G. Ramı́rez-Sánchez, and F.M. Tyers. 2011. Apertium: a free/open-source platform for rule-based machine translation. Machine Translation. Special Issue on Free/Open-Source Machine Translation, In press. find
7. J. Goodman and S. F. Chen. 1998. An empirical study of smoothing techniques for language modeling. Technical Report TR-10-98, Harvard University, August. find
8. W. J. Hutchins and H. L. Somers. 1992. An introduction to machine translation, volume 362. Academic Press New York. find
9. P. Koehn, H. Hoang, A. Birch, C. Callison-Burch, M. Federico, N. Bertoldi, B. Cowan, C. Shen, W.and Moran, R. Zens, C. Dyer, O. Bojar, A. Constantin, and E. Herbst. 2007. Moses: Open Source Toolkit for Statistical Machine Translation. In Proceedings of the 45th Annual Meeting of the ACL on Interactive Poster and Demonstration Sessions, pages 177–180. find
10. P. Koehn. 2005. Europarl: A parallel corpus for statistical machine translation. MT summit, 5:12–16. find
11. P. Koehn. 2010. Statistical Machine Translation. Cambridge University Press. find
12. A. Lopez and P. Resnik. 2006. Word-based alignment, phrase-based translation: What’s the link. In Proceedings of the 7th Conference of the Association for Machine Translation in the Americas, pages 90–99. find
13. F. J. Och and H. Ney. 2003. A systematic comparison of various statistical alignment models. Computational Linguistics, 29:19–51, March. find
14. F. J. Och. 2003. Minimum error rate training in statistical machine translation. In Proceedings of the 41st Annual Meeting of the ACL, pages 160–167. find
15. K. Papineni, S. Roukos, T. Ward, and W. Zhu. 2002. Bleu: a method for automatic evaluation of machine translation. In Proceedings of the 40th Annual Meeting of the ACL, pages 311–318. find our card
16. M. Popovic and H. Ney. 2006. Statistical machine translation with a small amount of bilingual training data. In LREC workshop on Minority Languages, pages 25– 29. find
17. H. Schwenk, S. Abdul-Rauf, L. Barrault, and J. Senellart. 2009. SMT and SPE machine translation systems for WMT’09. In Proceedings of the Fourth Workshop on Statistical Machine Translation, pages 130–134. find
18. A. Stolcke. 2002. SRILM-an extensible language modeling toolkit. In 7th International Conference on Spoken Language Processing, pages 901–904. find
19. G. Thurmair. 2009. Comparing different architectures of hybrid Machine Translation systems. In Proceedings MT Summit XII. find
20. F. M. Tyers. 2009. Rule-based augmentation of training data in Breton-French statistical machine translation. In Proceedings of the 13th Annual Conference of the European Association for Machine Translation, pages 213–217. find
21. Y. Zhang, S. Vogel, and A. Waibel. 2004. 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, pages 2051–2054. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/cartagena-2011-integrating-shallow-transfer-rules-into/