Elamur Research⌕
☰
home / papers / 2013 / Reconstructing an Indo-European Family Tree from Non-native English Texts

Reconstructing an Indo-European Family Tree from Non-native English Texts

Paper · 2013 · Ryo Nagata and Edward Whittaker

Summary

The paper asks how strongly a writer's first language shapes their English, and whether the relationships among first languages survive in that English. The authors test a bold hypothesis: if interference is strong enough, the Indo-European family tree should be recoverable from learner essays alone. They use ICLE v.2 essays by speakers of 11 Indo-European languages and filter out writers with mixed language backgrounds. Native (LOCNESS) and Japanese-authored (ICLE, NICE) English serve as reference sets. Each group's text is turned into token sequences in which content words become POS tags and function words stay as literal words, which reduces topic effects. Two models are then built and clustered with agglomerative hierarchical clustering: trigram language models compared with a symmetrized likelihood-ratio distance, and trigram frequency vectors compared with Euclidean distance. The resulting tree correctly separates the Italic, Germanic and Slavic branches. Analysis points to articles, NP construction and adverb placement as the main carriers of the signal. This matters for native language identification, error correction and historical linguistics.

Key points

  • Data: 11 Indo-European L1 subcorpora of ICLE v.2. Essays were kept only if the writer had a single native language that was also the home language and matched the subcorpus. References were LOCNESS (British and American students) and two Japanese-English sets (ICLE and NICE).
  • Representation: text is lowercased and POS-tagged. Content words become tags, proper nouns count as common nouns, and BOS/EOS markers are added. Determiners, prepositions, modals, pronouns, conjunctions, possessives, predeterminers and question adverbs are kept as words.
  • Model A uses trigram LMs with Kneser-Ney smoothing. The distance is a per-token log ratio of how well a corpus's own model fits it versus another group's model, averaged in both directions. Clustering uses average linkage. Model B uses Euclidean distance over relative trigram frequencies and gives similar trees that are easier to interpret.
  • Findings: the clusters match the Italic, Germanic and Slavic branches. The authors propose that interfamily distance > non-nativeness > intrafamily distance. A pilot run on Europarl English did not reproduce the canonical tree, though corpus heterogeneity may explain this.
  • The length of noun-noun compounds follows a Poisson distribution closely. For French-English, the predicted Pr(l=1) is 0.0303 and the observed value is 0.0299.

Interesting ideas

  • I1. Hide topic by masking content words with POS tags but keep function words literal, so the remaining sequence captures structural habits rather than subject matter.
  • I2. Measure how distant two text sources are by how much worse one source's model predicts the other's text, average both directions, then build a tree with hierarchical clustering.
  • I3. Check an unsupervised clustering against a known gold structure (the canonical family tree) and include outgroup and native control sets, so the result is interpretable rather than just plausible.
  • I4. Insert an explicit null-determiner symbol before bare common-noun NPs, so that a missing article becomes a visible token.
  • I5. Fit compound length with a simple Poisson distribution whose parameter is the average length, giving a cheap and interpretable prior.

For Mova

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

  • I1 (tree): Small function-word/POS trigram models could estimate how far apart treebanks or UD dialects are stylistically before deciding whether to merge or convert them.
  • I2 (learn): Clustering silver-data batches or witness-parser outputs this way would expose outlier sources that should be reviewed before training.
  • I3 (learn): Evaluations could likewise add negative-control corpora that should land far from the target cluster, confirming that a metric picks up the intended signal.
  • I4 (frames): An explicit "no determiner" marker would give coreference and entity introduction a clear cue for distinguishing new or generic mentions from definite, already-known ones.
  • I5 (tree): A Poisson prior over noun-noun compound chain length could flag implausibly long compound attachments in parser output for repair or review.

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

🎓Tokyo Institute of TechnologyTokyo · Japan22 works · 2,137 citations🎓Hyogo University of Teacher EducationKatō · Japan4 works · 117 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 Ryo Nagata and Edward Whittaker. 2013. Reconstructing an Indo-European Family Tree from Non-native English Texts. Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

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

@inproceedings{nagata2013reconstructing,
  title = {{Reconstructing an Indo-European Family Tree from Non-native English Texts}},
  author = {Ryo Nagata and Edward Whittaker},
  booktitle = {Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2013},
  url = {https://aclanthology.org/P13-1112/}
}

References

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

  1. Jan Aarts and Sylviane Granger, 1998. Tag sequences in learner corpora: a key to interlanguage grammar and discourse, pages 132–141. Longman, New York. find
  2. Bengt Altenberg and Marie Tapper, 1998. The use of adverbial connectors in advanced Swedish learners’ written English, pages 80–93. Longman, New York. find
  3. François Barbançon, Tandy Warnow, Steven N. Evans, Donald Ringe, and Luay Nakhleh. 2007. An experimental study comparing linguistic phylogenetic reconstruction methods. Statistics Technical Reports, page 732. find
  4. Vladimir Batagelj, Tomaž Pisanski, and Dami-jana Keržič. 1992. Automatic clustering of languages. Computational Linguistics, 18(3):339–352. find
  5. Robert S.P. Beekes. 2011. Comparative Indo-European Linguistics: An Introduction (2nd ed.). John Benjamins Publishing Company, Amsterdam. find
  6. Martin Chodorow, Michael Gamon, and Joel R. Tetreault. 2010. The utility of article and preposition error correction systems for English language learners: feedback and assessment. Language Testing, 27(3):419–436. ↗
  7. David Crystal. 1997. The Cambridge Encyclopedia of Language (2nd ed.). Cambridge University Press, Cambridge. find
  8. Niels Davidsen-Nielsen and Peter Harder, 2001. Speakers of Scandinavian languages: Danish, Norwegian, Swedish, pages 21–36. Cambridge University Press, Cambridge. find
  9. Alvar Ellegård. 1959. Statistical measurement of linguistic relationship. Language, 35(2):131–156. ↗
  10. Jessica Enright and Grzegorz Kondrak. 2011. The application of chordal graphs to inferring phylogenetic trees of languages. In Proc. of 5th International Joint Conference on Natural Language Processing, pages 8–13. find
  11. Sylviane Granger, Estelle Dagneaux, Fanny Meunier, and Magali Paquot. 2009. International Corpus of Learner English v2. Presses universitaires de Louvain, Louvain. find
  12. Russell D. Gray and Quentin D. Atkinson. 2003. Language-tree divergence times support the Anatolian theory of Indo-European origin. Nature, 426:435–438. ↗
  13. Jiawei Han and Micheline Kamber. 2006. Data Mining: Concepts and Techniques (2nd Ed.). Morgan Kaufmann Publishers, San Francisco. find
  14. Bing-Hwang Juang and Lawrence R. Rabiner. 1985. A probabilistic distance measure for hidden Markov models. AT&T Technical Journal, 64(2):391–408. ↗
  15. Kenji Kita. 1999. Automatic clustering of languages based on probabilistic models. Journal of Quantitative Linguistics, 6(2):167–171. ↗
  16. Reinhard Kneser and Hermann Ney. 1995. Improved backing-off for m-gram language modeling. In Proc. of International Conference on Acoustics, Speech, and Signal Processing, volume 1, pages 181–184. find
  17. Philipp Koehn. 2011. Europarl: A parallel corpus for statistical machine translation. In Proc. of 10th Machine Translation Summit, pages 79–86. find
  18. Moshe Koppel and Noam Ordan. 2011. Translationese and its dialects. In Proc. of 49th Annual Meeting of the Association for Computational Linguistics, pages 1318–1326.
  19. Moshe Koppel, Jonathan Schler, and Kfir Zigdon. 2005. Determining an author’s native language by mining a text for errors. In Proc. of 11th ACM SIGKDD International Conference on Knowledge Discovery in Data Mining, pages 624–628. find
  20. Alfred L. Kroeber and Charles D. Chriétien. 1937. Quantitative classification of Indo-European languages. Language, 13(2):83–103. ↗
  21. Ryo Nagata, Edward Whittaker, and Vera Sheinman. 2011. Creating a manually error-tagged and shallow-parsed learner corpus. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, pages 1210–1219. find
  22. Luay Nakhleh, Tandy Warnow, Don Ringe, and Steven N. Evans. 2005. A comparison of phylogenetic reconstruction methods on an Indo-European dataset. Transactions of the Philological Society, 103(2):171–192. ↗
  23. Taraka Rama and Anil Kumar Singh. 2009. From bag of languages to family trees from noisy corpus. In Proc. of Recent Advances in Natural Language Processing, pages 355–359. find
  24. Anna Giacalone Ramat and Paolo Ramat, 2006. The Indo-European Languages. Routledge, New York. find
  25. William Snyder. 1996. The acquisitional role of the syntax-morphology interface: Morphological compounds and syntactic complex predicates. In Proc. of Annual Boston University Conference on Language Development, volume 2, pages 728–735. find
  26. Masatoshi Sugiura, Masumi Narita, Tomomi Ishida, Tatsuya Sakaue, Remi Murao, and Kyoko Muraki. 2007. A discriminant analysis of non-native speakers and native speakers of English. In Proc. of Corpus Linguistics Conference CL2007, pages 84–89. find
  27. Michael Swan and Bernard Smith. 2001. Learner English (2nd Ed.). Cambridge University Press, Cambridge. find
  28. Hans van Halteren. 2008. Source language markers in EUROPARL translations. In Proc. of 22nd International Conference on Computational Linguistics, pages 937–944. find
  29. Sze-Meng J. Wong and Mark Dras. 2009. Contrastive analysis and native language identification. In Proc. Australasian Language Technology Workshop, pages 53–61. find
  30. Sze-Meng J. Wong, Mark Dras, and Mark Johnson. 2011. Exploiting parse structures for native language identification. In Proc. Conference on Empirical Methods in Natural Language Processing, pages 1600–1611. find
  31. Sze-Meng J. Wong, Mark Dras, and Mark Johnson. 2012. Exploring adaptor grammars for native language identification. In Proc. Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning, pages 699–709. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/nagata-2013-reconstructing-indo-european-family-tree/