Controlling Politeness in Neural Machine Translation via Side Constraints
By Rico Sennrich, Barry Haddow, Alexandra Birch. 2016
340 citations recorded by the source corpus.
Many languages use honorifics to express politeness, social distance, or the relative social status between the speaker and their ad-dressee(s). In machine translation from a language without honorifics such as English, it is difficult to predict the appropriate honorific, but users may want to control the level of politeness in the output. In this paper, we perform a pilot study to control honorifics in neural machine translation (NMT) via side constraints , focusing on English → German. We show that by marking up the (English) source side of the training data with a feature that en-codes the use of honorifics on the (German) target side, we can control the honorifics produced at test time. Experiments show that the choice of honorifics has a big impact on translation quality as measured by B LEU , and oracle experiments show that substantial im-provements are possible by constraining the translation to the desired level of politeness.
Subfields: Artificial Intelligence, Machine Translation
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Graham Neubig
- Yonatan Belinkov
- Mohit Bansal
- Luke S. Zettlemoyer
- Percy Liang
- Yoav Goldberg
- Jianfeng Gao
- Noah A. Smith
In Machine Translation
- Yonatan Belinkov
- Nadir Durrani
- Hassan Sajjad
- Philipp Koehn
- Fahim Dalvi
- Sudha Rao
- Markus Freitag
- Matteo Negri
Citations in the corpus
- WARP: Word-level Adversarial ReProgramming (2021, methodology): “In particular, multilingual neural machine translation models use special tokens in the input to control the target language (Ha et al., 2016; Johnson et al., 2017) or politeness of the translation (Sennrich et al., 2016).”
- Google’s Multilingual Neural Machine Translation System: Enabling Zero-Shot Translation (2016, background): “In Sennrich et al.,(2016a), the NMT framework has been extended to control the politeness level of the target translation by adding a special token to the source sentence.”
- Controlling Linguistic Style Aspects in Neural Language Generation (2017, background): “…of the generated text such as the identity of the speaker in a dialog setting (Li et al., 2016); the politeness of the generated text or the text length in a machine-translation setting (Sennrich et al., 2016a; Kikuchi et al., 2016); or the tense in generated movie reviews (Hu et al., 2017).”
- Adversarial Example Generation with Syntactically Controlled Paraphrase Networks (2018, methodology): “language with user-selected properties such as sentence length and formality but is incapable of generating paraphrases. Formachinetranslationoutput,Niuetal.(2017) control the level of formality whileSennrich et al. (2016) control the level of politeness. For dialogue,Li et al.(2016a) affect the out”
- Getting Gender Right in Neural Machine Translation (2019, methodology): “Our approach is similar to the work of Sennrich et al. (2016) on controlling politeness, where some sentence of the training data are followed with an ‘informal’ or ‘polite’ tag indicating the level of politeness expressed.”
- Controlling Output Length in Neural Encoder-Decoders (2016, background): “, 2016b), or politeness (Sennrich et al., 2016).”
- Style Transfer Through Back-Translation (2018, background): “…stylistic and demographic speaker traits in statistical machine translation (Rabinovich et al., 2016; Niu et al., 2017), and deep generative models controlling for a particular stylistic aspect, e.g., politeness (Sennrich et al., 2016), sentiment, or tense (Hu et al., 2017; Shen et al., 2017).”
- An Empirical Comparison of Domain Adaptation Methods for Neural Machine Translation (2017, methodology): “Tag based NMT has also been shown to be effective for control the politeness of translations (Sennrich et al., 2016a) and multilingual NMT (Johnson et al., 2016).”
- Extreme Adaptation for Personalized Neural Machine Translation (2018, methodology): “Table 3 shows final test scores for each model with statistical significance measured with paired boot- 5Recent NMT systems also commonly use sub-word units (Sennrich et al., 2016b).”
- Transfer Learning in Multilingual Neural Machine Translation with Dynamic Vocabulary (2018, methodology): “Inspired by [24], who automatically annotated the source side with artificial flags to manage the politeness level of the output, other works focused on controlling the grammatical voice [25], the text domain [26, 27], and enforcing gender agreement [28].”
- Controlling the Voice of a Sentence in Japanese-to-English Neural Machine Translation (2016, methodology): “It is the same model that Sennrich et al. (2016) used.”
- EDITOR: An Edit-Based Transformer with Repositioning for Neural Machine Translation with Soft Lexical Constraints (2020, background): “…tailored to spe-cific content or style constraints (Abu Sheikha and Inkpen, 2011; Mei et al., 2016), or via segment-level “side-constraints” (Sennrich et al., 2016a; Ficler and Goldberg, 2017; Scarton and Spe-cia, 2018), which condition generation on users’ stylistic preferences, but do not…”
- Domain Control for Neural Machine Translation (2016, methodology): “This method, though simple, has already proven to be effective to control the politeness level of a translation (Sennrich et al., 2016), or to support multi-lingual NMT models (Johnson et al.”
- DiaBLa: a corpus of bilingual spontaneous written dialogues for machine translation (2019, methodology): “The data is cleaned, tokenised and truecased using the Moses toolkit (Koehn et al., 2007) and tokens are split into subword units using BPE (Sennrich et al., 2016b).”
- Neural Machine Translation (2020, positive): “Sennrich et al. (2016d) split up all words to sub-word units, using character n-gram models and a segmentation based on the byte pair encoding compression algorithm.”
- Beyond Noise: Mitigating the Impact of Fine-grained Semantic Divergences on Neural Machine Translation (2021, positive): “DIV - TAGGED models distinguish equivalences from divergences by appending < EQ > vs. < DIV > tags as source-side constraints (Sennrich et al., 2016a).”
- Dear Sir or Madam, May I Introduce the GYAFC Dataset: Corpus, Benchmarks and Metrics for Formality Style Transfer (2018, methodology): “Under NMT models, Sennrich et al. (2016a) control the politeness of the translated text via side constraints, Niu et al. (2017) control the level of formality of MT output by selecting phrases of a requisite formality level from the k-best list during decoding.”