Elamur Research
☰
home / papers / 2016 / Controlling Politeness in Neural Machine Translation via Side Constraints

Controlling Politeness in Neural Machine Translation via Side Constraints

Written for the Research library (claude-opus-5-5) from the paper text. Original — aclanthology.org/N16-1005.pdf

Summary

Translating from English, which has no T-V distinction, into German forces the system to pick between the familiar (du) and the formal (Sie) form of address. The right choice usually cannot be inferred from the source text. The authors present a pilot study on English→German neural machine translation in which users choose the level of politeness at translation time.

They label German target sentences as informal, polite or neutral using rules over ParZu morphosyntactic analysis, which covers address pronouns and imperative verbs. The label is then attached as a special token (<T> or <V>) to the end of the English source sentence during training. The attentional encoder-decoder learns to attend to this "side constraint," so one model can produce either register on demand.

On OpenSubtitles test data, the constraints reliably steer the output toward the requested form. Choosing the form that matches the reference raises BLEU by about 3 points. This shows that pragmatic choices noticeably affect measured quality. The same trick could control other target-side features missing from the source, such as tense, gender, dialect or register.

Key points

  • Training used OpenSubtitles2012, with 5.58M sentence pairs. About 0.48M pairs were labelled polite and 1.09M informal, and the rest were left neutral. Test samples came from OpenSubtitles2013.
  • The model was a Groundhog attentional encoder-decoder with BPE subword units (vocabulary 90,000), trained for roughly 9 epochs (7 days).
  • Labelled sentences received their tag with probability α = 0.5. Neutral sentences received a random tag with the same probability. Tags were re-sampled every epoch so the model would learn to ignore irrelevant constraints and would not add unnecessary address pronouns.
  • On 2,000 test sentences whose English source contains "you," 96% of polite-constrained outputs were polite or neutral, and 98% of informal-constrained outputs were informal or neutral. The oracle constraint raised BLEU from 20.7 to 23.9. On a random 2,000-sentence sample, BLEU rose from 22.6 to 24.0.
  • The constraints act as soft preferences. Strong cues in the source can override them, as in "You foolish boy," which was rendered with du under every setting.

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: Linguistics & languages

Topics: Machine translation

Cited by

199 papers of our library cite it (of the 140673 whose bibliographies we read); 12 of them have a page here, 187 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/n16-1005

@inproceedings{sennrich2016controlling,
  title = {{Controlling Politeness in Neural Machine Translation via Side Constraints}},
  author = {Sennrich, Rico and Haddow, Barry and Birch, Alexandra},
  booktitle = {Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies},
  year = {2016},
  doi = {10.18653/v1/n16-1005},
  url = {https://doi.org/10.18653/v1/n16-1005}
}

References

19 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Walid Aransa, Holger Schwenk, and Loïc Barrault. 2015. Improving Continuous Space Language Models using Auxiliary Features. In Proceedings of the 12th International Workshop on Spoken Language Translation, pages 151–158, Da Nang, Vietnam. find
  2. Dzmitry Bahdanau, Kyunghyun Cho, and Yoshua Bengio. 2015. Neural Machine Translation by Jointly Learning to Align and Translate. In Proceedings of the International Conference on Learning Representations (ICLR). find our card
  3. Roger Brown and A. Gilman. 1960. The pronouns of power and solidarity. In T. Sebeok, editor, Style in Language. find
  4. The M.I.T. Press, Cambridge, MA. Kyunghyun Cho, Bart van Merrienboer, Caglar Gulcehre, Dzmitry Bahdanau, Fethi Bougares, Holger Schwenk, and Yoshua Bengio. 2014. Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation. In Proceedings of the 2014 Conference on Em… find
  5. Thierry Etchegoyhen, Lindsay Bywood, Mark Fishel, Panayota Georgakopoulou, Jie Jiang, Gerard Van Loenhout, Arantza Del Pozo, Mirjam Sepesy Maucec, Anja Turner, and Martin Volk. 2014. Machine Translation for Subtitling: A Large-Scale Evaluation. In Proceedings of the Ninth International Conference on… find
  6. Manaal Faruqui and Sebastian Pado. 2012. Towards a model of formal and informal address in English. In Proceedings of the 13th Conference of the European Chapter of the Association for Computational Linguistics, pages 623–633, Avignon, France. Association for Computational Linguistics. find
  7. Sébastien Jean, Kyunghyun Cho, Roland Memisevic, and Yoshua Bengio. 2015a. On Using Very Large Target Vocabulary for Neural Machine Translation. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language P… find our card
  8. Sébastien Jean, Orhan Firat, Kyunghyun Cho, Roland Memisevic, and Yoshua Bengio. 2015b. Montreal Neural Machine Translation Systems for WMT’15 . In Proceedings of the Tenth Workshop on Statistical Machine Translation, pages 134–140, Lisbon, Portugal. Association for Computational Linguistics. find
  9. Yangfeng Ji, Trevor Cohn, Lingpeng Kong, Chris Dyer, and Jacob Eisenstein. 2015. Document Context Language Models. ArXiv e-prints, November. find
  10. Minh-Thang Luong and Christopher D. Manning. 2015. Stanford Neural Machine Translation Systems for Spoken Language Domains. In Proceedings of the International Workshop on Spoken Language Translation 2015, Da Nang, Vietnam. find
  11. Tomas Mikolov and Geoffrey Zweig. 2012. Context dependent recurrent neural network language model. In 2012 IEEE Spoken Language Technology Workshop (SLT), pages 234–239, Miami, FL, USA. find
  12. Hideki Mima, Osamu Furuse, and Hitoshi Iida. 1997. Improving Performance of Transfer-driven Machine Translation with Extra-linguistic Information from Context, Situation and Environment. In Proceedings of the Fifteenth International Joint Conference on Ar-tifical Intelligence - Volume 2, IJCAI’97, p… find
  13. Morgan Kaufmann Publishers Inc. Shigeko Nariyama, Hiromi Nakaiwa, and Melanie Siegel. 2005. Annotating Honorifics Denoting Social Ranking of Referents. In Proceedings of the 6th International Workshop on Linguistically Interpreted Corpora (LINC-2005). find
  14. Doris Schüpbach, John Hajek, Jane Warren, Michael Clyne, Heinz-L. Kretzenbacher, and Catrin Norrby. 2006. A cross-linguistic comparison of address pronoun use in four European languages: Intralingual and interlingual dimensions . In Annual Meeting of the Australian Linguistic Society, Brisbane, Aust… find
  15. Rico Sennrich, Martin Volk, and Gerold Schneider. 2013. Exploiting Synergies Between Open Resources for German Dependency Parsing, POS-tagging, and Morphological Analysis. In Proceedings of the International Conference Recent Advances in Natural Language Processing 2013, pages 601–609, Hissar, Bulga… find
  16. Rico Sennrich, Barry Haddow, and Alexandra Birch. 2015. Neural Machine Translation of Rare Words with Subword Units. CoRR, abs/1508.07909. find
  17. SYSTRAN, 2004. SYSTRAN 5.0 User Guide. find
  18. Jörg Tiedemann. 2012. Parallel Data, Tools and Interfaces in OPUS. In Proceedings of the Eight International Conference on Language Resources and Evaluation (LREC’12), Istanbul, Turkey. find
  19. Tian Wang and Kyunghyun Cho. 2015. Larger-Context Language Modelling. ArXiv e-prints, November. find