Elamur Research⌕
☰
home / papers / 2009 / A semi-automatic part-of-speech tagging system for Middle English corpora: overcoming the challenges.

A semi-automatic part-of-speech tagging system for Middle English corpora: overcoming the challenges.

Paper · 2009 · Melania Sánchez Reed and Antonio Miranda García

Summary

The paper, by Sánchez Reed and Miranda García (University of Málaga), is a position and progress report on building a semi-automatic part-of-speech tagger for Middle English (ME). Modern English tagging is close to solved, but historical texts are harder. Spelling, capitalisation and punctuation were not standardised, dialects mixed, and Latin intrudes. Earlier efforts were limited:

  • Ruef's Old English tagger handled unknown words poorly.
  • The Texas system (Moon and Baldridge 2007) depended on modern parallel versions of the text.

The authors are working on the Malaga Corpus of Late Middle English scientific prose. They survey the linguistic and computational obstacles and propose a hybrid design. A tagger dictionary would supply candidate tags, and an HMM would choose among them using tag frequency and context. They also describe provisional fixes developed so far:

  • case-insensitive dictionary lookup
  • transcription-level markup for punctuation and multiword lemmas
  • fallback strategies for unknown words

No accuracy figures for their own system are given. Two problems remain open: handling the genitive and tagging unseen tokens. The work matters as groundwork for annotating historical corpora that have no modern parallel text.

Key points

  • Prior baselines:
    • Ruef's transformation-based OE tagger was trained on 108,000 words and reached 88.5% overall, with 91.5% on known tokens but only 56.5% on unknown ones.
    • Moon and Baldridge's Bible-alignment method combined a bigram tagger with the C&C maxent tagger and reached about 84%.
  • In the Texas study, a manually annotated set of 400–800 sentences beat their best bootstrapped tagger.
  • The Malaga Corpus edits ME manuscripts from Glasgow's Hunterian Collection. It holds about 250,000 words and aims for at least 500,000.
  • The planned pipeline has three steps:
    1. tokenization
    2. ambiguity look-up, where each token gets its candidate tags
    3. disambiguation, done by an HMM on top of a dictionary
  • An early markup scheme used ⟨**⟩, ⟨@@⟩ and ⟨%⟩ for punctuation status, ⟨ÑÑ⟩ for split lemmas, and >/< for discontinuous expressions. It proved too laborious and is being replaced by a simple sentence-boundary bar.

Interesting ideas

  • I1. Separate the possible tags of a word (dictionary look-up) from choosing one in context (statistical disambiguation), so each stage can be inspected and fixed on its own.
  • I2. When looking a word up, try the form exactly as written first, then with the opposite capitalisation, so that unreliable casing does not hide known entries.
  • I3. Mark each punctuation sign by its status (ignore, keep as modern, editor-inserted) instead of deleting or normalising it, so the original stays recoverable while the tool sees a clean segmentation.
  • I4. Explicitly join split or discontinuous multiword units into one lemma, with direction markers linking far-apart parts (e.g. not only … but also).
  • I5. Use a small hand-annotated seed: the reported finding that a few hundred annotated sentences outperformed bootstrapping is a useful sanity baseline for any self-training scheme.

For Mova

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

  • I1 (tokens): The en tagger could expose the candidate tag set per token before disambiguation, which makes tagging errors explainable as either "missing candidate" or "wrong choice".
  • I2 (tokens): A two-pass case lookup in the lexicon would make lemmatisation more robust on headlines, all-caps text and sloppy user input.
  • I3 (tokens): The tokenizer could carry a status flag on noisy punctuation, keeping sentence splitting explainable without losing the source text needed for generation.
  • I4 (tree): Such links could be treated as fixed or correlative units during UD conversion, so parsers and frames see one connective instead of two unrelated words.
  • I5 (learn): Silver-data pipelines from witness parsers should always be compared against a small gold-trained control, so that bootstrapping is known to actually beat 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.

📚 Shelves 5 Computer Science · 9,086 works Artificial Intelligence · 8,889 works Arts and Humanities · 9,296 works Language and Linguistics · 9,206 works Literature and Literary Theory · 340 works

🧭 Topics 3 English · 1,373 works Part-of-speech tagging · 91 works Historical linguistics & language change · 159 works

Cite

Cite as Melania Sánchez Reed and Antonio Miranda García. 2009. A semi-automatic part-of-speech tagging system for Middle English corpora: overcoming the challenges. SELIM 16.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.17811/selim.16.2009.121-147

@inproceedings{reed2009semi,
  title = {{A semi-automatic part-of-speech tagging system for Middle English corpora: overcoming the challenges.}},
  author = {Melania Sánchez Reed and Antonio Miranda García},
  booktitle = {SELIM 16},
  year = {2009},
  doi = {10.17811/selim.16.2009.121-147},
  url = {https://doi.org/10.17811/selim.16.2009.121-147}
}

References

29 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. AlGahtani, S., W. Black and J. McNaught 2009: Arabic Part-of-Speech Tagging Using Transformation-Based Learning. In K. Choukri and B. Maegaard, eds., Proceedings of the Second International Conference on Arabic Language Resources and Tools. (Cairo, Egypt): 66–70. find
  2. Altunyurt L., Z. Orhan and T. Güngör 2007: Towards Combining Rule-based and Statistical Part of Speech Tagging in Agglutinative Languages. Computer engineering 1.1: 66–69. find
  3. Selim 16 (2009) A semi-automatic POST system for ME corpora Brill, E. 1995: Unsupervised Learning of Disambiguation Rules for Part of Speech Tagging. Proceedings of the 3rd Workshop on Very Large Corpora. Massachusetts Institute of Technology (Cambridge, Massachusetts). 1–13. find
  4. Brill, E. 1992: A Simple Rule-Based Part of Speech Tagger. In M. Kauff man, ed., Proceedings of the DARPA. Speech and Natural Language Workshop. (San Mateo, California): 112–116. our card
  5. Calle-Martín, J. 2004: Punctuation Practice in a 15th-Century Arithmetical Treatise (MS Bodley 790). Neuphilologische Mitteilungen 4. 407–422. find
  6. Calle-Martín, J. and A. Miranda-García 2005: Editing Middle English Punctuation. The Case of MS Egerton 2622 (ff . 136–165). International Journal of English Studies 5. 27–44. find
  7. Cavella, C. and R. A. Kernodle 2003 [2010/08]: How the Past Aff ects the Future: The Story of the Apostrophe’. AU TESOL Working Papers 2. American University: Washington, DC. http://www1.american.edu/tesol/Working%20Papers/wpkernodlecavella.pdf ↗
  8. Charniak, E. 1997: Statistical Techniques for Natural Language Parsing. AI Magazine, 18.4. American Association for Artifi cial Intelligence. 33–43. find
  9. Daelemans W. and J. Zavrel 1996: MBT: A Memory-Based Part of Speech Tagger-Generator. Proceedings of the Fourth Workshop on Very Large Corpora (ACL SIGDAT). Copenhagen. 14–27. our card
  10. Fischl, W. 2009: Part of Speech Tagging – A Solved Problem? Unpublished work  om seminar/lecture. Technical University of Vienna: Austria. 1–8. find
  11. Giesbrecht, E. and S. Evert 2009: Is Part-of-Speech Tagging a Solved Task? An Evaluation of POS Taggers for the German Web as Corpus. Proceedings of the 5 th Web as Corpus Workshop (WAC5). Donostia. 27–35. find
  12. Kurath, H., S. Kuhn, J. Reidy and R. E. Lewis eds. 1956–2001 [2010/08]: Middle English Dictionary. Ann Arbour: University of Michigan Press. http://quod.lib.umich.edu/m/med/lookup.html ↗
  13. Lippmann, R. P. 1989: Review of Neural Networks for Speech Recognition. Neural Computation, Vol. i. MIT Press. 1–38. Cited in Schmid, H. 1994. Part-of-speech tagging with neural networks. Proceedings of the 15th Conference on Computational linguistics. Kyoto: Japan. 172–176. ↗
  14. Marqués-Aguado, T. 2009: Punctuation Practice in the Antidotary in G.U.L. MS Hunter 513 (ff . 37v–96v). Miscelánea 39. 55–72. find
  15. Selim 16 (2009) Melania Sánchez Reed & Antonio Miranda García McEnery, T. and A. Wilson 1997 [2010/08]: Corpus Linguistics. http://www.lancs.ac.uk/fss/courses/ling/corpus/contents.htm. ↗
  16. Miranda-García, A., J. Calle-Martín, D. Moreno-Olalla and J. L. Triviño-Rodríguez 2001: CALLOE: A Pedagogical Tool for the Learning of Old English. Old English Newsletter 34.3. 12–20. find
  17. Miranda-García, A., J. L. Triviño-Rodríguez and J. Calle-Martín 2000: A Morphological Analyzer of Old English Texts (MAOET). In A. M. Hornero and M. P. Navarro, eds. Proceedings of the 10th International Conference of SELIM. Zaragoza, Institución Fernando el Católico: 127– 145. find
  18. Moon, T. and J. Baldridge 2007: Part-of-Speech Tagging for Middle English through Alignment and Projection of Parallel Diachronic Texts. Proceedings of the 2007 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning (Prague, Czech Republic): 390–399. our card
  19. Moreno-Olalla, David and Miranda-García, Antonio 2009: An Annotated Corpus of Middle English Scientifi c Prose: aims & features. In Díaz Vera, Javier Enrique & Caballero Rodríguez, Rosario eds., Textual Healing: Studies in Medieval English Medical, Scientifi c and Technical Texts. Bern-New York, Peter Lang: 123–140. find
  20. Murata, M., Q. Ma and H. Isahara 2001: Part of Speech Tagging in Thai Language Using Support Vector Machine. Proceedings of Natural Language Processing and Neural Networks (NLPNN). Tokyo: Japan. 24–30. ↗
  21. Petti, A. G. 1977: English Literary Hands from Chaucer to Dryden. Harvard University Press: Cambridge Massachusetts. find
  22. Pianta, E. and R. Zanoli 2007: Tagpro: a System for Italian POS Tagging Based on SVM. Contributi Scientifi ci, 4.2. 8–9. find
  23. Ratnaparkhi, A. 1996: A Maximum Entropy Model for Part-of-Speech Tagging. Proceedings of the Empirical Methods in Natural Language Processing. Philadelphia, PA. 133–142. our card
  24. Ruef, B. [n. d.] [2010/08]: Transformation-Based Learning and Part-of-Speech Tagging of Old English. Zurich. https://fi les.ifi .uzh.ch/cl/gschneid/ KorpusSeminar/Beni_TBL_Slides.pdf Sánchez-Villamil, E., M. Forcada and R. Carrasco 2004: Unsupervised Training of a Finite-State Sliding-Window Part-of-Speech Tagger. In Vicedo, J. ↗
  25. Selim 16 (2009) A semi-automatic POST system for ME corpora L. et al. eds., Proceedings of the 4th International Conference España for Natural Language Processing (EsTAL) (Alicante, Spain): 454–463. find
  26. Schmid, H. 1994a: Probabilistic Part-of-Speech Tagging Using Decision Trees. Proceedings of International Conference on New Methods in Language Processing. (Manchester, UK): 44–49. find
  27. Schmid, H. 1994b: Part-of-Speech Tagging with Neural Networks. Proceedings of the 15th Conference on Computational linguistics (Kyoto, Japan): 172–176. find
  28. StatSo , Inc. 2010 [2010/08]: Electronic Statistics Textbook. Tulsa: StatSo . http://www.statso.com/textbook/ Van Guilder, L. 1995 [2010/08]: Automated Part of Speech Tagging: A Brief Overview. Handout for LING361. Georgetown University. http://ccl.pku. edu.cn/doubtfi re/NLP/Lexical_Analysis/Word_Segmentation_Tagging/ POS_Tagging_Overview/POS%20Tagging%20Overview.htm. ↗
  29. Received 28 Oct 2010; revision received 17 Nov 2010; accepted 22 Nov 2010 • Selim 16 (2009) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/reed-2019-semi-automatic-part-speech-tagging/