Elamur Research⌕
☰
home / papers / 2014 / Confusion Network for Arabic Name Disambiguation and Transliteration in Statistical Machine Translation

Confusion Network for Arabic Name Disambiguation and Transliteration in Statistical Machine Translation

Paper · 2014 · Young-Suk Lee

Summary

The paper addresses a common failure in Arabic-to-English statistical machine translation. Many Iraqi Arabic words can be read either as a name or as an ordinary word, such as SbAH, which means "Sabah" or "morning". Systems often mistranslate such names, especially when the training corpus barely contains the name reading. The author's pipeline has three parts:

  • Preprocessing: a maximum entropy POS tagger marks likely proper nouns with a $name class label.
  • Decoding: a phrase decoder takes a small confusion network that offers both the class label and the original word, and chooses between them by total decoding cost.
  • Postprocessing: a character-based phrase transliteration model spells out names left in Arabic script.

Translation and language models are trained on name-classed data from the DARPA TransTac corpus. Because the choice is probabilistic, the system can recover from tagging errors and needs no frequency threshold at decoding time. Human evaluation found a statistically significant quality gain on name-heavy test sets, with no loss on a set containing few names. The method is language-independent.

Key points

  • Training data: DARPA TransTac Iraqi Arabic–English, 766,410 sentence pairs (about 6.8M Arabic morpheme tokens and 7.3M English tokens), from military, medical, humanitarian and disaster-relief domains.
  • Name tagger: trained on the LDC Arabic Treebank (about 3M MSA morpheme tokens) plus about 63k in-house Iraqi tokens. Its NOUN_PROP F-score is about 93% on MSA names and about 81.4% on Iraqi names.
  • Classing rule: during training, only words seen fewer than 3 times and tagged NOUN_PROP are replaced by $name. This still gives 6,944 Arabic and 9,707 English $name tokens.
  • Language model: two 5-gram LMs with modified Kneser-Ney smoothing are interpolated, one without name classing and one with it (α = 0.1). The decoder uses 11 feature functions, and the tagger's −log probability is added to the translation costs.
  • Transliteration: exact match accuracy is 90% on 500 unique name pairs. Using the full decoder feature set instead of only TM+LM adds 15.6% absolute.

Interesting ideas

  • I1. Instead of committing to the tagger's name decision, pass both the class reading and the literal word forward as parallel paths, and let the global cost pick one; tagging errors become recoverable instead of fatal.
  • I2. Fold the upstream classifier's −log probability directly into the downstream scoring, so a confident tag pushes harder than a hesitant one.
  • I3. Apply class substitution only to rare words while training, which yields robust class statistics without erasing the senses of frequent words, but allow the class freely at inference.
  • I4. Treat target-side characters that have no source counterpart, here English vowels, as free units that attach to a neighbouring chunk during phrase extraction; this handles systematic asymmetry between writing systems.

For Mova

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

  • I1 (tokens): ambiguous capitalized words like "Bill", "Rose" or "May" could carry both PROPN and common-noun readings into the parser, with the tree score making the final call.
  • I2 (tree): tagger confidence could become an explicit, inspectable cost term in the perceptron parser's or gparse's decisions, so the explanation shows how much the tag drove each attachment.
  • I3 (learn): when building silver data, rare named entities could be abstracted into class tokens to train stable name-slot behaviour, while frequent words keep their ordinary lexical readings.
  • I4 (back): when generating or translating names through the meta-world, segments with no counterpart in the source could be attached to adjacent pieces rather than dropped or forced into alignment.

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, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

🧭 Topics 2Research subjects and languages the work is about.Machine translation · 1,105 worksSemitic languages · 3,066 works

Cite

Cite as Young-Suk Lee. 2014. Confusion Network for Arabic Name Disambiguation and Transliteration in Statistical Machine Translation. Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers.

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

@inproceedings{lee2014confusion,
  title = {{Confusion Network for Arabic Name Disambiguation and Transliteration in Statistical Machine Translation}},
  author = {Young-Suk Lee},
  booktitle = {Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers},
  year = {2014},
  url = {https://aclanthology.org/C14-1042/}
}

References

29 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Y. Al-Onaizan and K. Knight. 2002. Translating Named Entities Using Monolingual and Bilingual Resources. In Proceedings of the 40 th Annual Meeting of the Association for Computational Linguistics, pages 400–408.
  2. Y. Al-Onaizan and K. Knight. 2002. Machine Transliteration of Names in Arabic Text. In Proceedings of the Association for Computational Linguistics Workshop on Computational Approaches to Semitic Languages. find
  3. Y. Al-Onaizan and K. Papineni. 2006. Distortion models for Statistical Machine Translation. In Proceedings of the 21 st International Conference on Computational Linguistics and the 44 th Annual Meeting of the Association for Computational Linguistics, pages 529‒536. find
  4. N. Bertoldi, R. Zens, and M. Federico.2007. Speech translation by confusion network decoding. In Proceedings of the International Conference on Acoustics, Speech, and Signal Processing (ICASSP), pages 1297–1300. find
  5. P. Brown, S. Della Pietra, V. Della Pietra and R. Mercer. 1993. The mathematics of statistical machine translation: Parameter estimation. In Computational Linguistics, 19(2), pages 263‒311. find
  6. S. Chen and J. Goodman. 1998. An Empirical Study of Smoothing Techniques for Language Modeling. TR-10-98. Computer Science Group. Harvard University.
  7. C. Dyer, S. Muresan, and P. Resnik. 2008. Generalizing Word Lattice Translation. In Proceeding of the 46 th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, pages 1012– 1020.
  8. N. Habash and F. Sadat. 2006. Arabic Preprocessing Schemes for Statistical Machine Translation, In Proceedings of the North American Chapter of the Association for Computational Linguistics (NAACL), pages 49‒52.
  9. A. Hassan, H. Fahmy, and H. Hassan. 2007. Improving Named Entity Translation by Exploiting Comparable and Parallel Corpora. In Proceeding RANLP’07, pages 1–6. find
  10. U. Hermjakob, K. Knight, and H. Daume III. 2008. Name Translation in Statistical Machine Translation:Learning When to Transliterate. In Proceedings of the 46 th Annual Meeting of the Association for Computational Linguistics, pages 389–397. find
  11. S. Hewavitharana, D. Mehay, S. Ananthakrishnan, and P. Natarajan. 2013. Incremental Topic-Based Translation Model Adaptation for Conversational Spoken Language Translation. In Proceedings of the 51 st Annual Meeting of the Association for Computational Linguistics, pages 697-701.  : Cross-lingual Name Spell-and ing Normalization. In Proceedings of the Empirical Methods in Natural Language Processing, pages 391‒399. find
  12. F. Huang, A. Emami, and I. Zitouni. 2008. find
  13. When Harry Met Harri, P. Koehn, F. Josef Och, and D. Marcu. 2003. Statistical Phrase-Based Translation. In Proceedings of the 2003 Conference of the North American Chapter of the Association for Computational Linguistics on Human Language Technology – Volume 1, pages 127–133. find
  14. P. Koehn, H. Hoang, A. Birch, C. Callison-Burch, M. Federico, N. Bertoldi, B. Cowan, W. Shen, C. 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 45 th Annual Meeting of the Association of Computational Linguistics on Interactive Poster and Demonstration Sessions, pages 177‒180. find
  15. Y. Lee, K. Papineni, S. Roukos, O. Emam and H. Hassan. 2003. In Proceedings of the 41 st Annual Meeting of Association for Computational Linguistics ‒ Volume 1, pages 399‒406. find
  16. Y. Lee. 2004. Morphological Analysis for Statistical Machine Translation. In Proceedings of Human Language Technology Conference/North American Chapter of the Association for Computational Linguistics: Short Papers, pages 57‒60.
  17. H. Li, J. Zheng, H. Ji, Q. Li and W. Wang. 2013. Name-aware Machine Translation. In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pages 604–614. find
  18. F. Och and H. Ney. 2003. A Systematic Comparison of Various Statistical Alignment Models. In Computational Linguistics 29(1), pages 19‒51. MIT Press. find
  19. T. Onish, M. Utiyama and E. Sumita. 2010. Paraphrase Lattice for Statistical Machine Translation. In Proceedings of the 48 th Annual Meeting of the Association for Computational Linguistics Short Papers, pages 1–5. find
  20. K. Papineni, S. Roukos, T. Ward, and W. Zhu. 2002. BLEU: a Method for Automatic Evaluation of Machine Translation. In Proceedings of the 40 th Annual Meeting of the Association for Computational Linguistics, pages 311–318.
  21. A. Ratnaparkhi. 1996. A Maximum Entropy Model for Part-Of-Speech Tagging. In Proceedings of the Empirical Methods in Natural Language Processing, pages 133‒142.
  22. T. Sherif and G. Kondrak. 2007. Substring-Based Transliteration. In Proceedings of the 45 th Annual Meeting of the Association for Computational Linguistics, pages 944‒951. find
  23. B. G. Stalls and K. Knight. 1998. Translating Names and Technical Terms in Arabic Text. In Proceedings of the COLING/ACL Workshop on Computational Approaches to Semitic Languages.
  24. C. Tillmann. 2003. A Projection Extension Algorithm for Statistical Machine Translation. In Proceedings of the Empirical Methods in Natural Language Processing, pages 1‒8. find
  25. C. Tillmann and H. Ney. 2003. Word Reordering and a Dynamic Programming Beam-Search Algorithm for Statistical MT. Computational Linguistics 29(1), pages 97‒133. MIT Press. find
  26. S. Vogel, H. Ney and C. Tillmann. 1996. HMM-based word alignment in statistical machine translation. In Proceedings of the 16 th International Conference on Computational Linguistics, Volume 2, pages 836‒841. find
  27. O. Zayed, S. El-Beltagy, O. Haggag. An Approach for Extracting and Disambiguating Arabic Person’s Names Using Clustered Dictionaries and Scored Patterns. In Natural Language Processing and Information Systems Lecture Notes in Computer Science. Vol. 7934, 2013, pages 201‒212. find
  28. B. Zhao and S. Chen. 2009. A Simplex Armijo Downhill Algorithm for Optimizing Statistical Machine Translation Decoding. In Proceedings of Human Language Technology Conference/North American Chapter of the Association for Computational Linguistics Short Papers, pages 21‒24. find
  29. R. Zen and H. Ney. 2004. Improvements in phrase-based statistical machine translation. In Proceedings of Human Language Technology Conference/North American Chapter of the Association for Computational Linguistics, pages 257‒264.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/lee-2014-confusion-network-arabic-name-disambiguation/