Elamur Research⌕
☰
home / papers / 2017 / Low Resourced Machine Translation via Morpho-syntactic Modeling: The Case of Dialectal Arabic

Low Resourced Machine Translation via Morpho-syntactic Modeling: The Case of Dialectal Arabic

Paper · 2017 · Alexander Erdmann, Nizar Habash, Dima Taji et al.

Summary

The paper addresses machine translation between Arabic dialects, specifically Egyptian to Levantine. This direction has almost no naturally written parallel text and had been evaluated only once before. The authors built a 12,000-sentence multi-dialect corpus. Translators started from English BTEC sentences rather than from another dialect, which avoids "priming" the Levantine side with Egyptian phrasing.

On top of a small direct phrase-based model, they add several resources:

  • synthetic and phrase-pivot tables built through English, using the BOLT Egyptian–English and Levantine–English data;
  • monolingual AOC text;
  • morpho-syntactic constraints (definiteness, gender, number, POS) derived from MADAMIRA-EGY analyses.

On the blind test set, leaving the input untranslated scores 6.5 BLEU, the parallel-only system reaches 14.6, and pivoting plus morpho-syntactic modeling reaches 17.5. A manual error analysis shows larger gains than BLEU suggests. The work matters because it shows how to transfer tools and data from a better-resourced neighbor variety to a low-resource pair while limiting that neighbor's bias.

Key points

  • Data: BTEC Egy–Lev parallel data is split into 8,000 train, 500 tune, 1,500 dev and 2,000 test sentences. Pivot data is BOLT-egy (410,000 sentences) and BOLT-lev (180,000). Small monolingual sets come from AOC.

  • Baseline comparison (Table 2):

    ModelBLEUOOV
    No-translation6.48–
    Direct15.444.6%
    Synthetic16.75–
    Phrase pivot alone6.77–
    Dir+PP (direct + phrase pivot)17.410.9%

  • Preprocessing: MADAMIRA-EGY is also applied to Levantine, because no Levantine version exists. It is barred from orthographic rewrites so that Levantine words are not "Egyptianized." Segmentation follows D3 but is surface-oriented, so detokenization is plain concatenation.

  • Manual error analysis (100 dev sentences): adding the morpho-syntactic features reduces word errors by 20.2% and sentence errors by 13.5%. Fully correct sentences rise from 48 to 55.

  • Breakdown of the gains: errors involving only morpho-syntactic properties drop from 7.4% to 6.1% of tokens. Other errors drop even more, from 7.1% to 5.2% (about 26%).

Interesting ideas

  • I1. When a bigger sibling's analyzer is applied to a related variety, its normalization/rewrite step is switched off so it cannot impose the sibling's forms
  • I2. A segmentation scheme is chosen so that the original surface string is always recovered by simple concatenation, trading slight sparsity for guaranteed reversibility
  • I3. Agreement features such as definiteness, gender and number act as a feasibility filter that lets the model trust longer, rarer phrase units instead of falling back to short fragments
  • I4. Manual error analysis splits errors into the targeted category and everything else, and finds most of the gain arrived indirectly; BLEU understated the real improvement
  • I5. Parallel data is elicited from a neutral third language (English) rather than from one of the two varieties, to avoid translator priming

For Mova

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

  • I1 (tree): When converting or parsing non-standard English dialects with a standard-English tagger, restrict the tagger to annotation only, with no spelling normalization, so that dialect forms are not silently rewritten.
  • I2 (tokens): Make the tokenizer's splits strictly concatenation-reversible so every token maps exactly back to the source characters, which keeps explanations traceable.
  • I3 (back): During generation, use morphological agreement checks to license larger pre-built chunks over word-by-word assembly, improving fluency.
  • I4 (learn): Evaluations should break down errors by the category a feature targets versus side effects, so that indirect benefits and harms of a new rule become visible.
  • I5 (learn): When having an LLM teacher produce paraphrase or dialect variants, generate each from a neutral meaning description rather than from the other variant, to avoid copy bias in the silver data.

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

 Institutions

🎓The Ohio State UniversityColumbus · United States82 works · 18,502 citations🎓New York University Abu DhabiAbu Dhabi · United Arab Emirates122 works · 4,602 citations🎓Carnegie Mellon University QatarDoha · Qatar47 works · 2,773 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

🧭 Topics 3Research subjects and languages the work is about.Machine translation · 1,087 worksSemitic languages · 2,908 worksLow-resource languages · 267 works

Cite

Cite as Alexander Erdmann, Nizar Habash, Dima Taji and Houda Bouamor. 2017. Low Resourced Machine Translation via Morpho-syntactic Modeling: The Case of Dialectal Arabic. Proceedings of Machine Translation Summit XVI: Research Track.

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

@inproceedings{erdmann2017low,
  title = {{Low Resourced Machine Translation via Morpho-syntactic Modeling: The Case of Dialectal Arabic}},
  author = {Alexander Erdmann, Nizar Habash, Dima Taji and Houda Bouamor},
  booktitle = {Proceedings of Machine Translation Summit XVI: Research Track},
  year = {2017},
  url = {https://aclanthology.org/2017.mtsummit-papers.15/}
}

References

58 references, 28 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Al-Shargi, F., Kaplan, A., Eskander, R., Habash, N., and Rambow, O. (2016). Morphologically annotated corpora and morphological analyzers for Moroccan and Sanaani Yemeni Arabic. In 10th Language Resources and Evaluation Conference (LREC 2016).
  2. Almahairi, A., Cho, K., Habash, N., and Courville, A. (2016). First result on Arabic neural machine translation. arXiv preprint arXiv:1606.02680. ↗
  3. Armentano-Oller, C., Carrasco, R. C., Corbí-Bellot, A. M., Forcada, M. L., Ginestí-Rosell, M., Ortiz-Rojas, S., Pérez-Ortiz, J. A., Ramírez-Sánchez, G., Sánchez-Martínez, F., and Scalco, M. A. (2006). Open-source Portuguese–Spanish machine translation. In International Workshop on Computational Processing of the Portuguese Language, pages 50–59. find
  4. Bouamor, H., Habash, N., and Oflazer, K. (2014). A multidialectal parallel corpus of Arabic. In LREC, pages 1240–1245.
  5. Cettolo, M., Federico, M., and Bertoldi, N. (2010). Mining parallel fragments from comparable texts. In IWSLT, pages 227–234. find
  6. Chiang, D., Diab, M., Habash, N., Rambow, O., and Shareef, S. (2006). Parsing Arabic Dialects. In Proceedings of EACL, Trento, Italy. EACL. ↗
  7. Corbí Bellot, A. M., Forcada, M. L., Ortiz Rojas, S., Pérez-Ortiz, J. A., Ramírez Sánchez, G., Sánchez-Martínez, F., Alegría Loinaz, I., Mayor Martínez, A., Sarasola Gabiola, K., et al. (2005). An open-source shallow-transfer machine translation engine for the Romance languages of Spain. Proceedings of the 10th Conference of the European Association for Machine Translation, pages 79–86. find
  8. Cotterell, R. and Callison-Burch, C. (2014). A multi-dialect, multi-genre corpus of informal written Arabic. In LREC, pages 241–245.
  9. Diab, M., Habash, N., Rambow, O., Altantawy, M., and Benajiba, Y. (2010). Colaba: Arabic dialect annotation and processing. In LREC Workshop on Semitic Language Processing, pages 66–74. find
  10. Diab, Mona T and Al-Badrashiny, Mohamed and Aminian, Maryam and Attia, Mohammed and Elfardy, Heba and Habash, Nizar and Hawwari, Abdelati and Salloum, Wael and Dasigi, Pradeep and Eskander, Ramy. (2014). Tharwa: A Large Scale Dialectal Arabic-Standard Arabic-English Lexicon. In LREC, pages 3782–3789.
  11. Dong, M., Liu, Y., Luan, H.-B., Sun, M., Izuha, T., and Zhang, D. (2015). Iterative learning of parallel lexicons and phrases from non-parallel corpora. In IJCAI, pages 1250–1256. find
  12. Durrani, N., Al-Onaizan, Y., and Ittycheriah, A. (2014). Improving Egyptian-to-English SMT by mapping Egyptian into MSA. In International Conference on Intelligent Text Processing and Computational Linguistics, pages 271–282. find
  13. El Kholy, A. and Habash, N. (2011). Automatic error analysis for morphologically rich languages. In MT Summit XIII. find
  14. El Kholy, A. and Habash, N. (2012). Orthographic and morphological processing for English– Arabic statistical machine translation. Machine Translation, 26(1-2):25–45. ↗
  15. El Kholy, A. and Habash, N. (2015). Morphological Constraints for Phrase Pivot Statistical Machine Translation. In Machine Translation Summit, Miami. ↗
  16. El Kholy, A., Habash, N., Leusch, G., Matusov, E., and Sawaf, H. (2013). Language independent connectivity strength features for phrase pivot statistical machine translation. In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, volume 2, pages 412–418. find
  17. Elfardy, H. and Diab, M. (2013). Sentence Level Dialect Identification in Arabic. In Proceedings of the Association for Computational Linguistics, pages 456–461, Sofia, Bulgaria.
  18. Eskander, R., Habash, N., Rambow, O., and Pasha, A. (2016). Creating resources for Dialectal Arabic from a single annotation: A case study on Egyptian and Levantine. In Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, pages 3455–3465, Osaka, Japan.
  19. Eskander, R., Habash, N., Rambow, O., and Tomeh, N. (2013). Processing Spontaneous Orthography. In Proceedings of the 2013 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL-HLT), Atlanta, GA. find
  20. Farajian, M. A., Turchi, M., Negri, M., Bertoldi, N., and Federico, M. (2017). Neural vs. phrase-based machine translation in a multi-domain scenario. EACL 2017, page 280. find
  21. Habash, N., Diab, M., and Rambow, O. (2012a). Conventional Orthography for Dialectal Arabic. In Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC-2012), pages 711–718, Istanbul, Turkey.
  22. Habash, N., Eskander, R., and Hawwari, A. (2012b). A Morphological Analyzer for Egyptian Arabic. In Proceedings of the Twelfth Meeting of the Special Interest Group on Computational Morphology and Phonology, pages 1–9, Montréal, Canada.
  23. Habash, N. and Hu, J. (2009). Improving Arabic-Chinese Statistical Machine Translation using English as Pivot Language. In Proceedings of the Fourth Workshop on Statistical Machine Translation, pages 173–181, Athens, Greece. find
  24. Habash, N., Soudi, A., and Buckwalter, T. (2007). On Arabic Transliteration. In van den Bosch, A. and Soudi, A., editors, Arabic Computational Morphology: Knowledge-based and Empirical Methods. Springer. ↗
  25. Habash, N., Zalmout, N., Taji, D., Hoang, H., and Alzate, M. (2017). A parallel corpus for evaluating machine translation between arabic and european languages. EACL 2017, page 235.
  26. Habash, N. Y. (2010). Introduction to Arabic natural language processing, volume 3. Morgan & Claypool Publishers. find
  27. Hajič, J., Smrž, O., Zemánek, P., Šnaidauf, J., and Beška, E. (2004). Prague Arabic Dependency Treebank: Development in data and tools. In Proceedings of the NEMLAR International Conference on Arabic Language Resources and Tools, pages 110–117, Cairo, Egypt. find
  28. Heafield, K. (2011). KenLM: Faster and smaller language model queries. In Proceedings of the Sixth Workshop on Statistical Machine Translation, pages 187–197.
  29. Jarrar, M., Habash, N., Akra, D., and Zalmout, N. (2014). Building a corpus for Palestinian Arabic: a preliminary study. In Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing (ANLP), pages 18–27.
  30. Khalifa, S., Habash, N., Abdulrahim, D., and Hassan, S. (2016). A large scale corpus of Gulf Arabic. arXiv preprint arXiv:1609.02960. ↗
  31. Koehn, P. (2005). Europarl: A parallel corpus for statistical machine translation. In MT summit, volume 5, pages 79–86. find
  32. Koehn, P., Birch, A., and Steinberger, R. (2009). 462 Machine Translation Systems for Europe. In Proceedings of MT Summit, Ottawa, Canada. find
  33. Koehn, P., Hoang, H., Birch, A., Callison-Burch, C., Federico, M., Bertoldi, N., Cowan, B., Shen, W., Moran, C., Zens, R., et al. (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. ↗
  34. Luong, M.-T. and Manning, C. D. (2015). Stanford neural machine translation systems for spoken language domains. In Proceedings of the International Workshop on Spoken Language Translation. find
  35. Maamouri, M., Bies, A., Kulick, S., Ciul, M., Habash, N., and Eskander, R. (2014). Developing an Egyptian Arabic treebank: Impact of dialectal morphology on annotation and tool development. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC-2014). find
  36. Max, A., Bouamor, H., and Vilnat, A. (2012). Generalizing sub-sentential paraphrase acquisition across original signal type of text pairs. In Proceedings of the 2012 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 721–731. find
  37. McNeil, K. and Faiza, M. (2011). Tunisian Arabic corpus: Creating a written corpus of an unwritten language. In Workshop on Arabic Corpus Linguistics (WACL). find
  38. Meftouh, K., Harrat, S., Jamoussi, S., Abbas, M., and Smaili, K. (2015). Machine translation experiments on PADIC: A parallel Arabic dialect corpus. In The 29th Pacific Asia conference on language, information and computation.
  39. Munteanu, D. S. and Marcu, D. (2005). Improving machine translation performance by exploiting non-parallel corpora. Computational Linguistics, 31(4):477–504. ↗
  40. Muraki, K. (1987). PIVOT: Two-phase machine translation system. In MT Summit Manuscripts and Program, pages 81–83. find
  41. Och, F. J. and Ney, H. (2003). A systematic comparison of various statistical alignment models. Computational linguistics, 29(1):19–51. ↗
  42. Papineni, K., Roukos, S., Ward, T., and Zhu, W.-J. (2002). BLEU: a method for automatic evaluation of machine translation. In Proceedings of the 40th annual meeting on association for computational linguistics, pages 311–318.
  43. Pasha, A., Al-Badrashiny, M., Kholy, A. E., Eskander, R., Diab, M., Habash, N., Pooleery, M., Rambow, O., and Roth, R. (2014). MADAMIRA: A Fast, Comprehensive Tool for Morphological Analysis and Disambiguation of Arabic. In Proceedings of LREC, Reykjavik, Iceland.
  44. Resnik, P. and Smith, N. A. (2003). The web as a parallel corpus. Computational Linguistics, 29(3):349–380. ↗
  45. Sadat, F. and Habash, N. (2006). Combination of Arabic preprocessing schemes for statistical machine translation. In Proceedings of the 21st International Conference on Computational Linguistics and the 44th annual meeting of the Association for Computational Linguistics, pages 1–8.
  46. Sajjad, H., Darwish, K., and Belinkov, Y. (2013). Translating dialectal Arabic to English. In Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers), pages 1–6, Sofia, Bulgaria. find
  47. Salloum, W. and Habash, N. (2011). Dialectal to standard arabic paraphrasing to improve arabic-english statistical machine translation. In Proceedings of the first workshop on algorithms and resources for modelling of dialects and language varieties, pages 10–21. ↗
  48. Salloum, W. and Habash, N. (2012). Elissa: A Dialectal to Standard Arabic Machine Translation System. In Proceedings of the 24th International Conference on Computational Linguistics (COLING 2012): Demonstration Papers, pages 385–392, Mumbai, India.
  49. Sawaf, H. (2010). Arabic dialect handling in hybrid machine translation. In Proceedings of AMTA, Denver, Colorado.
  50. Schwenk, H. and Senellart, J. (2009). Translation model adaptation for an Arabic/French news translation system by lightly-supervised training. In MT Summit. find
  51. Tachicart, R. and Bouzoubaa, K. (2014). A hybrid approach to translate Moroccan Arabic dialect. In Intelligent systems: Theories and applications (sita-14), 2014 9th international conference on, pages 1–5. IEEE. find
  52. Takezawa, T., Sumita, E., Sugaya, F., Yamamoto, H., and Yamamoto, S. (2002). Toward a broad-coverage bilingual corpus for speech translation of travel conversations in the real world. In LREC, pages 147–152.
  53. Tyers, F. M., Font, H. A. i., Fronteddu, G., and Martín-Mor, A. (2017). Rule-based machine translation for the Italian-Sardinian language pair. In The Prague Bulletin of Mathematical Linguistics No. 108, pages 221–232. ↗
  54. Utiyama, M. and Isahara, H. (2007). A comparison of pivot methods for phrase-based statistical machine translation. In HLT-NAACL, pages 484–491.
  55. Wu, H. and Wang, H. (2007). Pivot language approach for phrase-based statistical machine translation. In Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics, pages 856–863, Prague, Czech Republic. ↗
  56. Zaidan, O. F. and Callison-Burch, C. (2011). The Arabic online commentary dataset: an annotated dataset of informal Arabic with high dialectal content. In Proceedings of the Association for Computational Linguistics, Portland, Oregon, USA.
  57. Zbib, R., Malchiodi, E., Devlin, J., Stallard, D., Matsoukas, S., Schwartz, R., Makhoul, J., Zaidan, O. F., and Callison-Burch, C. (2012). Machine translation of Arabic dialects. In Proceedings of the 2012 conference of the North American chapter of the association for computational linguistics: Human language technologies, pages 49–59.
  58. Zhang, D., Kim, J., Crego, J. M., and Senellart, J. (2016). Boosting neural machine translation. CoRR, abs/1612.06138. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/erdmann-2017-low-resourced-machine-translation-morpho/