Elamur Research⌕
☰
home / papers / 2014 / Biber Redux: Reconsidering Dimensions of Variation in American English

Biber Redux: Reconsidering Dimensions of Variation in American English

Paper · 2014 · Rebecca J. Passonneau, Nancy Ide, Songqiao Su et al.

Summary

Genre-aware models help parsing, speech recognition, translation and retrieval. However, most genre classifiers treat genres as separate bins. This paper re-tests Biber's 1988 claim that texts instead vary continuously along several dimensions. Biber found these dimensions by applying PCA to British English. The authors repeat the analysis on MASC, a heterogeneous corpus of American English from the 1990s onward. MASC also covers newer kinds of text: email, blogs, tweets and collaboratively written ficlets.

They use a compact set of fewer than three dozen features. Each feature was extracted with freely available NLP tools and checked by hand, and the set includes Named Entity features Biber lacked. PCA yields four main components. Three resemble Biber's dimensions, and the fourth is driven by the entity features. Six genre groups, built by hierarchical clustering, separate well on some components but overlap on others. The authors measure this with mean Bhattacharyya distance. This partial separability may explain why genre classification results differ so much across studies. The resulting space could also be used to compare corpora, for example when adapting parsers to a new domain.

Key points

  • Biber's original study ran PCA on about one million words of British English, drawn from the LOB corpus (1961), the London-Lund spoken corpus and his own letters. He reported six dimensions, such as involved vs. informational and narrative vs. non-narrative.
  • This study uses MASC (Manually Annotated Sub-Corpus), whose validated annotations allow features that were hard to obtain in the 1980s. The features are fewer than three dozen and were manually validated.
  • PCA gives four major components. Three echo Biber's dimensions, and the fourth comes from Named Entity features. For example, the fourth component separates expository from offline-interactive texts, which overlap heavily on the first.
  • Early experiments with k-means, Naïve Bayes and PCA produced unstable classification results. PCA explained why: genre regions were distinct on one axis but overlapping on another.
  • In the Bhattacharyya distance table, every genre except Discursive Text separates clearly from at least one other genre, with values near or above 0.80. Discursive Text is therefore predicted to be the hardest to classify.

Interesting ideas

  • I1. Treat genre as a position in a continuous space with several axes rather than a single label, and ask on which axis two classes actually differ. This makes classifier failures diagnosable instead of mysterious.
  • I2. Score how separable classes are, using mean Bhattacharyya distance in the reduced space, before building a classifier for them. This gives an early warning that some distinctions cannot be learned from the chosen features.
  • I3. Measure corpus similarity in a low-dimensional space of interpretable stylistic features rather than by lexical overlap. The aim is to predict which training data best fits a new target domain.
  • I4. Use a small set of hand-validated features from existing tools instead of hundreds of noisy ones. Each resulting dimension then stays explainable through its feature loadings.
  • I5. Predict where an unseen genre will fall from the situation in which it is used. The paper's example is chats patterning with face-to-face talk.

For Mova

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

  • I1 (tokens): Placing each input text in a small register space could tell the tagger and lemmatizer when they face tweet-like or conversational text and should switch settings or flag lower confidence.
  • I2 (learn): Before training any genre-conditioned component, compute class separability on the projected features and abandon distinctions that show no distance, as a kind of negative control.
  • I3 (tree): Projecting a new text collection into the genre space could pick which treebank or parser model (perceptron vs. gparse) to apply, or which silver data to add for domain adaptation.
  • I4 (world): Register axes built from readable features such as entity density could be recorded as a property of the text's world, signalling how much named-entity tracking a fable needs compared with a news report.
  • I5 (back): Generation could aim for a target point in the genre space, for instance a conversational register, and check the output's projected scores to confirm it reached the intended style.

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

🎓Columbia UniversityNew York · United States241 works · 54,053 citations🎓Vassar CollegeTown of Poughkeepsie · United States6 works · 1,118 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 Rebecca J. Passonneau, Nancy Ide, Songqiao Su and Jesse Stuart. 2014. Biber Redux: Reconsidering Dimensions of Variation in American English. Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers.

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

@inproceedings{passonneau2014biber,
  title = {{Biber Redux: Reconsidering Dimensions of Variation in American English}},
  author = {Rebecca J. Passonneau, Nancy Ide, Songqiao Su and Jesse Stuart},
  booktitle = {Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers},
  year = {2014},
  url = {https://aclanthology.org/C14-1054/}
}

References

30 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Douglas Biber. 1988. Variation across Speech and Writing. Cambridge University Press, Cambridge, UK. ↗
  2. Douglas Biber. 1993. The multi-dimensional approach to linguistic analyses of genre variation: An overview of methodology and findings. Computers and the Humanities, 26:331–345. ↗
  3. Ivan Bulyko and Mari Ostendorf. 2003. Getting more mileage from web text sources for conversational speech language modeling using class-dependent mixtures. In Proc. HLT-NAACL 2003, pages 7–9. find
  4. Eugene Charniak. 2000. A maximum-entropy-inspired parser. In Proceedings of the 1st North American Chapter of the Association for Computational Linguistics Conference, NAACL 2000, pages 132–139, Stroudsburg, PA, USA. Association for Computational Linguistics.
  5. Guy Barrett Coleman and Harry C Andrews. 1979. Image segmentation by clustering. Proceedings of the IEEE, 67(5):773–785. ↗
  6. Sergey Feldman, Marius Marin, Julie Medero, and Mari Ostendorf. 2009. Classifying factored genres with part-of-speech histograms. In Proceedings of Human Language Technologies: The 2009 Annual Conference of the North American Chapter of the Association for Computational Linguistics, Companion Volume: Short Papers, pages 173–176, Boulder, Colorado, June. Association for Computational Linguistics. find
  7. Stefan Th. Gries, John Newman, Cyrus Shaoul, and Philip Dilts. 2009. N-grams and the clustering of genres. Paper presented at the workshop on Corpus, Colligation, Register Variation at the 31st Annual Meeting of the Deutsche Gesellschaft fr Sprachwissenschaft. find
  8. Nancy Ide, Collin Baker, Christiane Fellbaum, and Rebecca Passonneau. 2010. The Manually Annotated Sub-Corpus: A Community Resource for and by the People. In Proceedings of the ACL 2010 Conference Short Papers, pages 68–73, Uppsala, Sweden, July. Association for Computational Linguistics. find
  9. Rukmini Iyer and Mari Ostendorf. 1999. Relevance weighting for combining multi-domain data for n-gram language modeling. Computer Speech & Language, 13(3):267–282. find
  10. Thomas Kailath. 1967. The divergence and Bhattacharyya distance measures in signal selection. Communication Technology, IEEE Transactions on, 15(1):52–60. ↗
  11. Ioannis Kanaris and Efstathios Stamatatos. 2009. Learning to recognize webpage genres. Information Processing and Management, 45(5):499–512, September. ↗
  12. Jussi Karlgren and Douglass Cutting. 1994. Recognizing text genres with simple metrics using discriminant analysis. In Proceedings of the 15th Conference on Computational Linguistics - Volume 2, COLING ’94, pages 1071–1075, Stroudsburg, PA, USA. Association for Computational Linguistics. find
  13. Brett Kessler, Geoffrey Nunberg, and Hinrich Schütze. 1997. Automatic detection of text genre. In Proceedings of the 35th Annual Meeting of the Association for Computational Linguistics and Eighth Conference of the European Chapter of the Association for Computational Linguistics, ACL ’98, pages 32–38, Stroudsburg, PA, USA. Association for Computational Linguistics. ↗
  14. Adam Kilgarriff. 2001. Comparing corpora. International Journal of Corpus Linguistics, 6(1):1–37. find
  15. Teuvo Kohonen. 1995. Self-organizing Maps. Springer-Verlag, Berlin. ↗
  16. Yong-Bae Lee and Sung Hyon Myaeng. 2002. Text genre classification with genre-revealing and subject-revealing features. In SIGIR ’02: Proceedings of the 25th annual international ACM SIGIR conference on Research and development in information retrieval, pages 145–150, New York, NY, USA. ACM Press. find
  17. David Martinez and Eneko Agirre. 2000. One sense per collocation and genre/topic variations. In Proceedings of the 2000 Joint SIGDAT Conference on Empirical Methods in Natural Language Processing and Very Large Corpora: Held in Conjunction with the 38th Annual Meeting of the Association for Computational Linguistics - Volume 13, EMNLP ’00, pages 207–215, Stroudsburg, PA, USA. Association for Computational Linguistics. find
  18. David McClosky, Eugene Charniak, and Mark Johnson. 2010. Automatic domain adaptation for parsing. In Human Language Technologies: The 2010 Annual Conference of the North American Chapter of the Association for Computational Linguistics, HLT ’10, pages 28–36, Stroudsburg, PA, USA. Association for Computational Linguistics.
  19. A. Mehler, S. Sharoff, and M. Santini. 2010. Genres on the Web: Computational Models and Empirical Studies. Text, Speech and Language Technology. Springer. find
  20. Kishore Papineni, Salim Roukos, Todd Ward, and Wei-Jing Zhu. 2002. BLEU: a method for automatic evaluation of machine translation. In Proceedings of 40th Annual Meeting of the Association for Computational Linguistics, pages 311–318, Philadelphia, Pennsylvania, USA, July. Association for Computational Linguistics.
  21. Andreas Rauber and Alexander Müller-Kögler. 2001. Integrating automatic genre analysis into digital libraries. In First ACM-IEEE Joint Conference on Digital Libraries, pages 1–10. find
  22. Sujith Ravi, Kevin Knight, and Radu Soricut. 2008. Automatic prediction of parser accuracy. In Proceedings of the 2008 Conference on Empirical Methods in Natural Language Processing, pages 887–896, Honolulu, Hawaii, October. Association for Computational Linguistics. find
  23. Joseph Le Roux, Jennifer Foster, Joachim Wagner, Rasul Samad, Zadeh Kaljahi, and Anton Bryl. 2012. DUC-Paris13 systems for the SANCL 2012 shared task. find
  24. Marina Santini. 2004. A shallow approach to syntactic feature extraction for genre classification. Technical Report ITRI-04-02, Information Technology Research Institute, University of Brighton. Also published in Proceedings of the 7th Annual Colloquium for the UK Special Interest Group for Computational Linguistics, Birmingham, UK. find
  25. Ruhi Sarikaya, Agustı́n Gravano, and Yuqing Gao. 2005. Rapid language model development using external resources for new spoken dialog domains. In International Congress of Acoustics, Speech, and Signal Processing (ICASSP), pages 573–576, Philadelphia, PA, USA. IEEE, Signal Processing Society. find
  26. Serge Sharoff, Zhili Wu, and Katja Markert. 2010. The web library of Babel: evaluating genre collections. In Nicoletta Calzolari (Conference Chair), Khalid Choukri, Bente Maegaard, Joseph Mariani, Jan Odijk, Stelios Piperidis, Mike Rosner, and Daniel Tapias, editors, Proceedings of the Seventh International Conference on Language Resources and Evaluation (LREC’10), Valletta, Malta, may. European Language Resources Association (ELRA). find
  27. Efstathios Stamatatos, Nikos Fakotakis, and George Kokkinakis. 2000a. Text genre detection using common word frequencies. In Proceedings of the 18th Conference on Computational Linguistics - Volume 2, COLING ’00, pages 808–814, Stroudsburg, PA, USA. Association for Computational Linguistics. find
  28. Efstathios Stamatatos, George Kokkinakis, and Nikos Fakotakis. 2000b. Automatic text categorization in terms of genre and author. Computational Linguistics, 26(4):471–495, December. ↗
  29. Wei Wang, Klaus Macherey, Wolfgang Macherey, Franz Och, and Peng Xu. 2012. Improved domain adaptation for statistical machine translation. In AMTA-2012. find
  30. Nigel G. Ward and Steven D. Werner. 2013. Using dialog-activity similarity for spoken information retrieval. In Frédéric Bimbot, Christophe Cerisara, Cécile Fougeron, Guillaume Gravier, Lori Lamel, François Pellegrino, and Pascal Perrier, editors, 14th Annual Conference of the International Speech Communication Association, Interspeech, pages 1569–1573. ISCA. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/passonneau-2014-biber-redux-reconsidering-dimensions-variation/