Elamur Research⌕
☰
home / papers / 2007 / Chinese Syntactic Reordering for Statistical Machine Translation

Chinese Syntactic Reordering for Statistical Machine Translation

Summary

Chinese and English order constituents differently. Relative clauses, prepositional phrases and temporal expressions often come before the head in Chinese but after it in English. Phrase-based SMT systems handle such long-range movement poorly. The authors parse Chinese sentences with a Treebank-trained parser and apply hand-written transformation rules to the trees. The rules were designed by examining phrase categories in the Penn Chinese Treebank guidelines, and they rearrange the source so it follows English order more closely. The reordered text is used as preprocessing for both training and test data of the MOSES phrase-based system. The rules target three categories: verb phrases, noun phrases and localizer phrases. On NIST 2006 evaluation data, BLEU rose from 28.52 to 30.86. The authors also manually checked how often sentences were reordered correctly and measured each rule group's contribution. The work matters because it shows that explicit source-side syntax helps translation even when Chinese parsing, which depends on word segmentation, is fairly error-prone.

Key points

  • The rules move several kinds of pre-verbal modifiers after the sibling VP:
    • PPs and LCPs (localizer phrases);
    • temporal NPs, identified by containing at least one NT-tagged word;
    • QPs expressing duration or frequency.
  • Inside NPs, only DNP modifiers (built as "XP+DEG") and clausal CP modifiers, such as relative clauses, are candidates for reordering. ADJP, ADVP, DP, QP and PP have their own internal order and are left unchanged.
  • VP rules helped more than NP rules on the development set, even though NP rules fire more often. LC rules applied alone gave a surprisingly small gain.
  • The baseline system, trained on unreordered data, still scored higher on reordered dev input than on original input (32.19 vs. 31.57). This is a 0.62 BLEU gain from reordering only the test side.
  • An earlier NIST 2006 submission used a subset of the rules on top of a lexicalized-reordering system and gained 1.34 BLEU. The current setup with an unlexicalized model gains 1.82.

Interesting ideas

  • I1. Reorder the source as a preprocessing step. This keeps the decoder untouched and works with any phrase-based system. It is an easy way to add syntax for a language pair with systematic word-order differences.
  • I2. Choose rules by frequency and systematic contrast. The authors skipped rare constructions, such as topicalized objects, because the parser is unreliable on them and their effect would be small. This is a sensible cost–benefit filter for any hand-crafted rule set.
  • I3. Use existing tags as cheap triggers. The NT tag for temporal nouns made it easy to detect temporal NPs. For locatives the authors deliberately did not use NR, since proper nouns are not reliably places. The lesson is to check whether an annotation scheme already encodes the distinction a rule needs.
  • I4. Diagnose with matched and mismatched input. Testing both systems on original and reordered inputs separates the effect of input order from the quality of the learned phrase table. The test was inconclusive here, but the design carries over to other preprocessing studies.
  • I5. Output reordering lattices instead of a single reordered string. The authors propose this as future work to make the pipeline more robust to parse and rule errors. It is a general remedy for error propagation in pipelines.

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

Fields: Machine translation · Natural language processing

Topics: Machine translation

Cited by

104 papers of our library cite it (of the 140673 whose bibliographies we read); 3 of them have a page here, 101 not yet.

Cite

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

@inproceedings{wang2007chinese,
  title = {{Chinese Syntactic Reordering for Statistical Machine Translation}},
  author = {Wang, Chao and Collins, Michael and Koehn, Philipp},
  booktitle = {Proceedings of the 2007 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning (EMNLP-CoNLL)},
  year = {2007},
  url = {https://aclanthology.org/D07-1077/}
}

References

22 references, 8 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Adam L. Berger, Stephen A. Della Pietra, and Vincent J. Della Pietra. 1996. A maximum entropy approach to natural language processing. Computational Linguistics, 22(1):39–69. ↗ our card
  2. Peter F. Brown, Stephen A. Della Pietra, Vincent J. Della Pietra, John D. Lafferty, and Robert L. Mercer. 1992. Analysis, statistical transfer, and synthesis in machine translation. In Proceedings of Conference on Theoretical and Methodological Issues in Machine Translation. find
  3. David Chiang. 2005. A hierarchical phrase-based model for statistical machine translation. In Proceedings of ACL, pages 263–270. find our card
  4. Michael Collins, Philipp Koehn, and Ivona Kučerová. 2005. Clause restructuring for statistical machine translation. In Poceedings of ACL, pages 531–540. find our card
  5. Michael Collins. 1997. Three generative, lexicalized models for statistical parsing. In Proceedings of ACL. find
  6. Michael Collins. 2002. Discriminative training methods for hidden Markov models: Theory and experiments with perceptron algorithms. In Proceedings of EMNLP. find our card
  7. Yuan Ding and Martha Palmer. 2005. Machine translation using probablistic synchronous dependency insertion grammars. In Proceedings of ACL, pages 541– 548, Ann Arbor, Michigan. find
  8. Liang Huang, Kevin Knight, and Aravind Joshi. 2006. Statistical syntax-directed translation with extended domain of locality. In Proceedings of AMTA. find our card
  9. Philipp Koehn, Amittai Axelrod, Alexandra Birch Mayne, and Chris Callison-Burch. 2005. Edinburgh system description. In IWSLT Speech Translation Evaluation. find
  10. Philipp Koehn, Hieu Hoang, Alexandra Birch, Chris Callison-Burch, Marcello Federico, Nicola Bertoldi, Brooke Cowan, Wade Shen, Christine Moran, Richard Zens, Chris Dyer, Ondrej Bojar, Alexandra Con-strantin, and Evan Herbst. 2007. Moses: Open source toolkit for statistical machine translation. In Proceedings of ACL, Demonstration Session. find
  11. Shankar Kumar and William Byrne. 2005. Local phrase reordering models for statistical machine translation. In Proceedings of HLT-EMNLP. find
  12. Dekang Lin. 2004. A path-based transfer model for machine translation. In Proceedings of Coling 2004, pages 625–630, Geneva, Switzerland, Aug 23–Aug 27. COLING. find
  13. Yang Liu, Qun Liu, and Shouxun Lin. 2006. Tree-tostring alignment template for statistical machine translation. In Proceedings of ACL, pages 609–616. find
  14. Daniel Marcu, Wei Wang, Abdessamad Echihabi, and Kevin Knight. 2006. SPMT: Statistical machine translation with syntactified target language phrases. In Proceedings of EMNLP, pages 44–52, Sydney, Australia. find
  15. Sonja Niessen and Hermann Ney. 2004. Statistical machine translation with scarce resources using morphosyntactic information. Computational Linguistics, 30(2):181–204. find
  16. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. BLEU: a method for automatic evaluation of machine translation. In Proceedings of ACL. find our card
  17. Chris Quirk, Arul Menezes, and Colin Cherry. 2005. Dependency treelet translation: Syntactically informed phrasal SMT. In Proceedings of ACL, pages 271–279, Ann Arbor, Michigan. find our card
  18. Honglin Sun and Daniel Jurafsky. 2004. Shallow semantic parsing of Chinese. In Proceedings of NAACL-HLT. find
  19. Christoph Tillmann. 2004. A block orientation model for statistical machine translation. In Proceedings of HLT-NAACL, Boston, MA, USA. find
  20. Fei Xia and Michael McCord. 2004. Improving a statistical MT system with automatically learned rewrite patterns. In Proceedings of COLING. find
  21. Nianwen Xue, Fei Xia, Fu-Dong Chiou, and Martha Palmer. 2005. The Penn Chinese Treebank: Phrase structure annotation of a large corpus. Natural Language Engineering, 11(2):207–238. find
  22. Kenji Yamada and Kevin Knight. 2001. A syntax-based statistical translation model. In Proceedings of ACL. find our card

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wang-2007-chinese-syntactic-reordering-statistical-machine/