Elamur Research⌕
☰
home / papers / 2011 / “I Thou Thee, Thou Traitor”: Predicting Formal vs. Informal Address in English Literature

“I Thou Thee, Thou Traitor”: Predicting Formal vs. Informal Address in English Literature

Paper · 2011 · Manaal Faruqui and Sebastian Padó

Summary

Modern English uses a single "you" for both familiar and polite address. Many target languages (German, French, Russian) must choose between a T and a V form, so this social information has to be inferred when translating. Faruqui and Padó test whether the T/V distinction can be recovered from English text alone.

They built a parallel English–German corpus of 115 nineteenth-century novels from Project Gutenberg. German second-person pronouns were labeled T or V with four rules based on capitalization and sentence position. These labels were then projected onto aligned English sentences, which produced training data without any manual annotation.

In a human study, two annotators labeled sentences more accurately and agreed more often when they could see surrounding sentences. A Naive Bayes classifier using only words from the current sentence reached 67% accuracy. That is roughly the level of humans judging isolated sentences, but still below humans who had context.

The work frames T/V as one case of a broader cross-lingual problem: information the source language leaves implicit but the target language requires. It also releases a corpus for studying that problem.

Key points

  • Corpus construction:
    • The corpus covers 115 novels with about 0.5M sentences per language.
    • Preprocessing used the EUROPARL splitter/tokenizer, Gargantuan for sentence alignment (it allows one-to-many links), Giza++ for word alignment and TreeTagger for lemmas and POS. No parser was used.
  • Label coverage:
    • The German rules label only somewhat over half of the pronouns, about 15% of sentences (45K V, 30K T).
    • Fewer than 1% of labeled German sentences contain both T and V.
    • The English side receives 53K V and 35K T sentences, split into 72 training novels (63K sentences) and 21 test novels (15K sentences).
  • Human study (100 sentences):
    • Without context, annotator accuracy was 63/65% and inter-annotator agreement 68%.
    • With three sentences before and after, accuracy rose to 70/69% and agreement to 81%.
    • Both annotators were native speakers of languages that mark T/V.
  • Classifier results:
    • Baselines: random 50%, majority class V 60.1%.
    • Naive Bayes with words: 66.1%; words+POS bigrams: 65.0%; words+archaic class (25 archaic verbs and pronouns): 67.0%.
    • Only the gap between the best and worst model is significant by a χ² test.
  • Most indicative words:
    • V cues include titles (madam, gentlemen, monsieur) and formulaic "permit".
    • T cues include thee and didst.
    • Many top cues are simply character names from individual novels, such as Fogg and Mowgli.

Interesting ideas

  • I1. Free labels from translations: A distinction that one language marks overtly can label a parallel text in a language that hides it, which yields training data at no annotation cost.
  • I2. Coarser projection unit: The authors projected labels to whole sentences rather than through pronoun word alignments, because those alignments were unreliable; this trades granularity for robustness.
  • I3. Conservative labeling rules: The disambiguation rules deliberately leave ambiguous cases unlabeled, accepting low coverage in exchange for cleaner labels, because a later classifier generalizes to unseen sentences anyway.
  • I4. Human ceilings by context size: Measuring human performance both on isolated sentences and with a small surrounding window separates what is locally recoverable from what needs discourse. Inspecting the top features also exposed character-name leakage.
  • I5. Global relationship constraint: The authors propose that a known social relation between two characters should fix the form of address across all their exchanges, which beats sentence-level guessing.

For Mova

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

  • I1 (learn): German or Ukrainian translations of English tales could act as an extra witness that produces silver labels for addressee formality or the speaker–hearer relation.
  • I2 (learn): when projecting any cross-lingual signal into silver data, choose the coarsest unit the alignment can support, then refine it with the UD tree.
  • I3 (learn): silver-data agreement rules can be allowed to abstain, so that Mova trains only on confident instances.
  • I4 (learn): evaluations should report sentence-only and context-window ceilings, and could add negative controls that mask proper names to catch per-story memorization.
  • I5 (world): store the speaker–addressee relation (rank, intimacy) as a state in the world of the text, so that address choice and politeness inferences, and later T/V choice in translation, can draw on it.

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

🎓Heidelberg UniversityHeidelberg · Germany38 works · 41,356 citations🎓Indian Institute of Technology KharagpurKharagpur · India24 works · 714 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 1Research subjects and languages the work is about.English · 2,263 works

Cite

Cite as Manaal Faruqui and Sebastian Padó. 2011. “I Thou Thee, Thou Traitor”: Predicting Formal vs. Informal Address in English Literature. Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies.

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

@inproceedings{faruqui2011i,
  title = {{“I Thou Thee, Thou Traitor”: Predicting Formal vs. Informal Address in English Literature}},
  author = {Manaal Faruqui and Sebastian Padó},
  booktitle = {Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies},
  year = {2011},
  url = {https://aclanthology.org/P11-2082/}
}

References

19 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. John Ardila. 2003. (Non-Deictic, Socio-Expressive) T-/V-Pronoun Distinction in Spanish/English Formal Locutionary Acts. Forum for Modern Language Studies, 39(1):74–86. ↗
  2. John A. Bateman. 1988. Aspects of clause politeness in japanese: An extended inquiry semantics treatment. In Proceedings of the 26th Annual Meeting of the Association for Computational Linguistics, pages 147–154, Buffalo, New York. find
  3. Fabienne Braune and Alexander Fraser. 2010. Improved unsupervised sentence alignment for symmetrical and asymmetrical parallel corpora. In Coling 2010: Posters, pages 81–89, Beijing, China. find
  4. Roger Brown and Albert Gilman. 1960. The pronouns of power and solidarity. In Thomas A. Sebeok, editor, Style in Language, pages 253–277. MIT Press, Cambridge, MA. find
  5. Alexander Clark. 2003. Combining distributional and morphological information for part of speech induction. In Proceedings of the Conference of the European Chapter of the Association for Computational Linguistics, pages 59–66, Budapest, Hungary.
  6. Pedro Domingos and Michael J. Pazzani. 1997. On the optimality of the simple Bayesian classifier under zero-one loss. Machine Learning, 29(2–3):103–130. ↗
  7. David Elson, Nicholas Dames, and Kathleen McKeown. 2010. Extracting social networks from literary fiction. In Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, pages 138–147, Uppsala, Sweden. ↗
  8. Alexander Fraser. 2009. Experiments in morphosyntactic processing for translating to and from German. In Proceedings of the Fourth Workshop on Statistical Machine Translation, pages 115–119, Athens, Greece. find
  9. Jerry Hobbs and Megumi Kameyama. 1990. Translation by abduction. In Proceedings of the 13th International Conference on Computational Linguistics, Helsinki, Finland. ↗
  10. Hiroshi Kanayama. 2003. Paraphrasing rules for automatic evaluation of translation into japanese. In Proceedings of the Second International Workshop on Paraphrasing, pages 88–93, Sapporo, Japan. find
  11. Philipp Koehn. 2005. Europarl: A Parallel Corpus for Statistical Machine Translation. In Proceedings of the 10th Machine Translation Summit, pages 79–86, Phuket, Thailand. find
  12. Heinz L. Kretzenbacher, Michael Clyne, and Doris Schüpbach. 2006. Pronominal Address in German: Rules, Anarchy and Embarrassment Potential. Australian Review of Applied Linguistics, 39(2):17.1–17.18. ↗
  13. Alexander Künzli. 2010. Address pronouns as a problem in French-Swedish translation and translation revision. Babel, 55(4):364–380. ↗
  14. Zhifei Li and David Yarowsky. 2008. Mining and modeling relations between formal and informal Chinese phrases from web corpora. In Proceedings of the 2008 Conference on Empirical Methods in Natural Language Processing, pages 1031–1040, Honolulu, Hawaii. find
  15. Franz Josef Och and Hermann Ney. 2003. A Systematic Comparison of Various Statistical Alignment Models. Computational Linguistics, 29(1):19–51. ↗
  16. Michael Schiehlen. 1998. Learning tense translation from bilingual corpora. In Proceedings of the 36th Annual Meeting of the Association for Computational Linguistics and 17th International Conference on Computational Linguistics, pages 1183–1187, Montreal, Canada. find
  17. Helmut Schmid. 1994. Probabilistic Part-of-Speech Tagging Using Decision Trees. In Proceedings of the International Conference on New Methods in Language Processing, pages 44–49. find
  18. Doris Schüpbach, John Hajek, Jane Warren, Michael Clyne, Heinz Kretzenbacher, and Catrin Norrby. 2006. A cross-linguistic comparison of address pronoun use in four European languages: Intralingual and interlingual dimensions. In Proceedings of the Annual Meeting of the Australian Linguistic Society, Brisbane, Australia. find
  19. David Yarowsky and Grace Ngai. 2001. Inducing multilingual POS taggers and NP bracketers via robust projection across aligned corpora. In Proceedings of the 2nd Meeting of the North American Chapter of the Association of Computational Linguistics, pages 200–207, Pittsburgh, PA. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/faruqui-2011-i-thou-thee-thou-traitor/