Skip to main content
SpringerLink
Log in

Statistical Machine Translation and Turkish

  • Chapter
  • First Online:
Turkish Natural Language Processing

Abstract

Machine translation is one of the most important applications of natural language processing. The last 25 years have seen tremendous progress in machine translation, enabled by the development of statistical techniques and availability of large-scale parallel sentence corpora from which statistical models of translation can be learned. Turkish poses quite many challenges for statistical machine translation as alluded to in Chap. 1, owing mainly to its complex morphology. This chapter discusses in more detail the challenges of Turkish in the context of statistical machine translation and describes two widely different approaches that have been employed in the last several years to English to Turkish machine translation.

This is a preview of subscription content, log in via an institution to check access.

Access this chapter

Log in via an institution
Chapter
USD 29.95
Price excludes VAT (USA)
  • Available as PDF
  • Read on any device
  • Instant download
  • Own it forever
eBook
USD 84.99
Price excludes VAT (USA)
  • Available as EPUB and PDF
  • Read on any device
  • Instant download
  • Own it forever
Softcover Book
USD 109.99
Price excludes VAT (USA)
  • Compact, lightweight edition
  • Dispatched in 3 to 5 business days
  • Free shipping worldwide - see info
Hardcover Book
USD 109.99
Price excludes VAT (USA)
  • Durable hardcover edition
  • Dispatched in 3 to 5 business days
  • Free shipping worldwide - see info

Tax calculation will be finalised at checkout

Purchases are for personal use only

Institutional subscriptions

Notes

  1. 1.

    www.statmt.org/wmt16/ (Accessed Sept. 14, 2017).

  2. 2.

    International Workshop on Spoken Language Translation: workshop2013.iwslt.org/ (Accessed Sept. 14, 2017).

  3. 3.

    Note that on the English side, the filler for [something] would come in the middle of this phrase.

  4. 4.

    See Chap. 1 for details.

  5. 5.

    This disambiguator has about 94% accuracy.

  6. 6.

    Ideally, it would have been very desirable to actually do derivational morphological analysis on the English side, so that one could, for example, analyze accession into access plus a marker indicating nominalization.

  7. 7.

    The training set in the first row of Table 10.2 was limited to sentences on the Turkish side which had at most 90 tokens (roots and bound morphemes) in total in order to comply with the limitations of the GIZA++ alignment tool. However when only the content words are included, we have more sentences to include since much less number of sentences violate the length restriction when morphemes/function words are removed.

  8. 8.

    It should be noted that what to selectively attach to the root should be considered on a per-language basis; if Turkish were to be aligned with a language with similar morphological markers, this perhaps would not have been needed.

  9. 9.

    Using the content word data improved performance for all representations except the baseline.

  10. 10.

    We ran MERT on the baseline model and the morphologically segmented models forcing -weight-d to range a very small around 0.1, but letting the other parameters range in their suggested ranges. Even though the procedure came back claiming that it achieved a better BLEU score on the tune set, running the new model on the test set did not show any improvement at all. This may have been due to the fact that the initial choice of -weight-d along with -dl set to -1 provides such a drastic improvement that perturbations in the other parameters do not have much impact.

  11. 11.

    We arrived at this combination by experimenting with the decoder to avoid the almost monotonic translation we were getting with the default parameters. These parameters boosted the BLEU scores substantially compared to default parameters used by the decoder.

  12. 12.

    We should also note that all sentences were lowercased so that we would not have to deal with exact capitalization issue at that stage.

  13. 13.

    The meanings of various tags are as follows: Dependency Labels: PMOD—Preposition Modifier; POS—Possessive. Part-of-Speech Tags for the English words: +IN—Preposition; +PRP$—Possessive Pronoun; +JJ—Adjective; +NN—Noun; +NNS—Plural Noun. Morphological Feature Tags in the Turkish Sentence: +A3pl—3rd person plural; +P3sg—3rd person singular possessive; +Loc—Locative case. Note that we mark an English plural noun as +NN_NNS to indicate that the root is a noun and there is a plural morpheme on it. Note also that economic is also related to relations but we are not interested in such content words and their relations.

  14. 14.

    We use _ to prefix such syntactic tags on the English side.

  15. 15.

    The order is important in that we would like to attach the same sequence of function words in the same order so that the resulting tags on the English side are the same.

  16. 16.

    We outline two additional rules later when we see a more complex example in Fig. 10.4.

  17. 17.

    For example, the morphological analyzer outputs +A3sg to mark a singular noun, if there is no explicit plural morpheme. Such markers are removed.

  18. 18.

    The tune set was not used in this work but reserved for future work so that meaningful comparisons could be made.

  19. 19.

    It is possible that the ten test sets are not mutually exclusive.

  20. 20.

    These allow and do not penalize unlimited distortions, but increase decoding time.

  21. 21.

    In Moses, factors are separated by a ‘|’ symbol.

  22. 22.

    Concatenating Root and Tags gives the Surface form, in that the surface is unique given this concatenation.

  23. 23.

    Note that for Turkish, this representation is equivalent to surface words in that the surface is unique given this representation.

  24. 24.

    Note that in this case, the translations would be generated in the same format, but we then split such postpositions from the words they are attached to, during decoding, and then evaluate the BLEU score.

  25. 25.

    In order to provide a simple and clear representation, the example sentences contain the surface form of the words as opposed to the morphemic representation used earlier.

  26. 26.

    For instance, consider the example in Fig. 10.4 involving if with some additional modifiers added to the intervening noun phrase.

References

  • Bisazza A, Federico M (2009) Morphological pre-processing for Turkish to English statistical machine translation. In: Proceedings of IWSLT, Tokyo, pp 129–135

    Google Scholar 

  • Carpuat M (2009) Toward using morphology in French-English phrase-based SMT. In: Proceedings of WMT, Athens, pp 150–154

    Google Scholar 

  • Chung T, Gildea D (2009) Unsupervised tokenization for machine translation. In: Proceedings of EMNLP, Singapore, pp 718–726

    Google Scholar 

  • Corston-Oliver S, Gamon M (2004) Normalizing German and English inflectional morphology to improve statistical word alignment. In: Proceedings of AMTA, Washington, DC

    Google Scholar 

  • Durgar-El Kahlout İ (2009) A prototype English-Turkish statistical machine translation system. PhD thesis, Sabancı University, Istanbul

    Google Scholar 

  • Durgar-El Kahlout İ, Oflazer K (2006) Initial explorations in English to Turkish statistical machine translation. In: Proceedings of WMT, New York, NY, pp 7–14

    Google Scholar 

  • Durgar-El Kahlout İ, Oflazer K (2010) Exploiting morphology and local word reordering in English-to-Turkish phrase-based statistical machine translation. IEEE Trans Audio Speech Lang Process 18(6):1313–1322

    Google Scholar 

  • Durgar-El Kahlout İ, Mermer C, Doğan MU (2012) Recent improvements in statistical machine translation between Turkish and English. In: Vertan C, von Hahn W (eds) Multilingual processing in Eastern and Southern EU languages: low-resourced technologies and translation. Cambridge Scholars Publishing, Cambridge

    Google Scholar 

  • Eyigöz E, Gildea D, Oflazer K (2013a) Multi-rate HMMs for word alignment. In: Proceedings of WMT, Sofia, pp 494–502

    Google Scholar 

  • Eyigöz E, Gildea D, Oflazer K (2013b) Simultaneous word-morpheme alignment for statistical machine translation. In: Proceedings of NAACL-HLT, Atlanta, GA, pp 32–40

    Google Scholar 

  • Goldwater S, McClosky D (2005) Improving statistical MT through morphological analysis. In: Proceedings of EMNLP, Vancouver, BC, pp 676–683

    Google Scholar 

  • Koehn P, Och FJ, Marcu D (2003) Statistical phrase-based translation. In: Proceedings of NAACL-HLT, Edmonton, AB, pp 127–133

    Google Scholar 

  • Koehn P, Hoang H, Birch A, Callison-Burch C, Federico M, Bertoldi N, Cowan B, Shen W, Moran C, Zens R, Dyer C, Bojar O, Constantin A, Herbst E (2007) Moses: open source toolkit for statistical machine translation. In: Proceedings of ACL, Prague, pp 177–180

    Google Scholar 

  • Lee YS (2004) Morphological analysis for statistical machine translation. In: Proceedings of NAACL-HLT, Boston, MA, pp 57–60

    Google Scholar 

  • Luong MT, Nakov P, Kan MY (2010) A hybrid morpheme-word representation for machine translation of morphologically rich languages. In: Proceedings of EMNLP, Cambridge, MA, pp 148–157

    Google Scholar 

  • Mermer C, Akın AA (2010) Unsupervised search for the optimal segmentation for statistical machine translation. In: Proceedings of the ACL student research workshop, Uppsala, pp 31–36

    Google Scholar 

  • Minkov E, Toutanova K, Suzuki H (2007) Generating complex morphology for machine translation. In: Proceedings of ACL, Prague, pp 128–135

    Google Scholar 

  • Naradowsky J, Toutanova K (2011) Unsupervised bilingual morpheme segmentation and alignment with context-rich hidden semi-markov models. In: Proceedings of ACL-HLT, Portland, OR, pp 895–904

    Google Scholar 

  • Nguyen T, Vogel S, Smith NA (2010) Nonparametric word segmentation for machine translation. In: Proceedings of COLING, Beijing, pp 815–823

    Google Scholar 

  • Niessen S, Ney H (2004) Statistical machine translation with scarce resources using morpho-syntatic information. Comput Linguist 30(2):181–204

    Google Scholar 

  • Nivre J, Hall J, Nilsson J, Chanev A, Eryiğit G, Kübler S, Marinov S, Marsi E (2007) MaltParser: a language-independent system for data-driven dependency parsing. Nat Lang Eng 13(2):95–135

    Google Scholar 

  • Oflazer K (1994) Two-level description of Turkish morphology. Lit Linguist Comput 9(2):137–148

    Google Scholar 

  • Oflazer K (1996) Error-tolerant finite-state recognition with applications to morphological analysis and spelling correction. Comput Linguist 22(1):73–99

    Google Scholar 

  • Oflazer K, Durgar-El Kahlout İ (2007) Exploring different representational units in English-to-Turkish statistical machine translation. In: Proceedings of WMT, Prague, pp 25–32

    Google Scholar 

  • Papineni K, Roukos S, Ward T, Zhu WJ (2002) BLEU: a method for automatic evaluation of machine translation. In: Proceedings of ACL, Philadelphia, PA, pp 311–318

    Google Scholar 

  • Popovic M, Ney H (2004) Towards the use of word stems and suffixes for statistical machine translation. In: Proceedings of LREC, Lisbon, pp 1585–1588

    Google Scholar 

  • Sadat F, Habash N (2006) Combination of Arabic preprocessing schemes for statistical machine translation. In: Proceedings of COLING-ACL, Sydney, pp 1–8

    Google Scholar 

  • Schmid H (1994) Probabilistic part-of-speech tagging using decision trees. In: Proceedings of the international conference on new methods in language processing, Manchester

    Google Scholar 

  • Stolcke A (2002) SRILM – an extensible language modeling toolkit. In: Proceedings of ICSLP, Denver, CO, vol 2, pp 901–904

    Google Scholar 

  • Talbot D, Osborne M (2006) Modelling lexical redundancy for machine translation. In: Proceedings of COLING-ACL, Sydney, pp 969–976

    Google Scholar 

  • Tantuğ AC, Oflazer K, Durgar-El Kahlout İ (2008) BLEU+: a tool for fine-grained BLEU computation. In: Proceedings of LREC, Marrakesh, pp 1493–1499

    Google Scholar 

  • Toutanova K, Klein D, Manning CD, Singer Y (2003) Feature-rich part-of-speech tagging with a cyclic dependency network. In: Proceedings of NAACL-HLT, Edmonton, AB, pp 252–259

    Google Scholar 

  • Yang M, Kirchhoff K (2006) Phrase-based backoff models for machine translation of highly inflected languages. In: Proceedings of EACL, Trento, pp 41–48

    Google Scholar 

  • Yeniterzi R (2009) Syntax-to-morphology alignment and constituent reordering in factored phrase-based statistical machine translation from English to Turkish. Master’s thesis, Sabancı University, Istanbul

    Google Scholar 

  • Yeniterzi R, Oflazer K (2010) Syntax-to-morphology mapping in factored phrase-based statistical machine translation from English to Turkish. In: Proceedings of ACL, Uppsala, pp 454–464

    Google Scholar 

  • Yılmaz E, Durgar-El Kahlout İ (2014) The use of recurrent neural networks language model in Turkish-English machine translation. In: Proceedings of IEEE signal processing and communications applications conference, Trabzon, pp 1247–1250

    Google Scholar 

  • Yılmaz E, Durgar-El Kahlout İ, Aydın B, Özil ZS (2013) TÜBİTAK Turkish-English submissions for IWSLT 2013. In: Proceedings of IWSLT, Heidelberg, pp 152–159

    Google Scholar 

  • Yuret D, Türe F (2006) Learning morphological disambiguation rules for Turkish. In: Proceedings of NAACL-HLT, New York, NY, pp 328–334

    Google Scholar 

  • Zollmann A, Venugopal A, Vogel S (2006) Bridging the inflection morphology gap for Arabic statistical machine translation. In: Proceedings of NAACL-HLT, New York, NY, pp 201–204

    Google Scholar 

Download references

Author information

Authors and Affiliations

  1. Carnegie Mellon University Qatar, Doha-Education City, Qatar

    Kemal Oflazer

  2. Özyeǧin University, Istanbul, Turkey

    Reyyan Yeniterzi

  3. TÜBİTAK-BİLGEM, Gebze, Kocaeli, Turkey

    İlknur Durgar-El Kahlout

Authors
  1. Kemal Oflazer
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Reyyan Yeniterzi
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. İlknur Durgar-El Kahlout
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Kemal Oflazer .

Editor information

Editors and Affiliations

  1. Carnegie Mellon University Qatar, Doha-Education City, Qatar

    Kemal Oflazer

  2. Electrical and Electronic Engineering, Boğaziçi University, Istanbul-Bebek, Turkey

    Murat Saraçlar

Rights and permissions

Reprints and permissions

Copyright information

© 2018 Springer International Publishing AG, part of Springer Nature

About this chapter

Check for updates. Verify currency and authenticity via CrossMark

Cite this chapter

Oflazer, K., Yeniterzi, R., Kahlout, İ.DE. (2018). Statistical Machine Translation and Turkish. In: Oflazer, K., Saraçlar, M. (eds) Turkish Natural Language Processing. Theory and Applications of Natural Language Processing. Springer, Cham. https://doi.org/10.1007/978-3-319-90165-7_10

Download citation

  • DOI: https://doi.org/10.1007/978-3-319-90165-7_10

  • Published:

  • Publisher Name: Springer, Cham

  • Print ISBN: 978-3-319-90163-3

  • Online ISBN: 978-3-319-90165-7

  • eBook Packages: Computer ScienceComputer Science (R0)

Publish with us

Policies and ethics

Search

Navigation