Elamur Research⌕
☰
home / papers / 2016 / Stylometric Analysis of Early Modern Period English Plays

Stylometric Analysis of Early Modern Period English Plays

Paper · 2016 · Mark Eisen, Santiago Segarra, Gabriel Egan et al.

Summary

Many English plays from the late 16th and early 17th centuries have uncertain authorship, because title pages were often wrong and co-writing went unrecorded. The authors use word adjacency networks (WANs) built from function words to study this drama. For each play, they count how often one function word is followed by another within a 10-word window inside the same speech, and nearer pairs count more (decay 0.75). Normalising each row of counts turns the network into a Markov chain. The chains of an author's secure plays are pooled into a profile, and a disputed text is assigned to the profile with the lowest relative entropy. On the undisputed plays of six playwrights, the method attributes 92.6% correctly and beats the frequency-only methods used in earlier Shakespeare studies. It also handles collaborative plays. Act-by-act and scene-by-scene runs on disputed works, such as Arden of Faversham, Edward III and Titus Andronicus, confirm accepted divisions and point to some new ones. The main lesson is that the order and spacing of grammatical words carry an authorial signal that plain counts miss.

Key points

  • Settings: α = 0.75 and D = 10, taken from earlier work. The text units are individual speeches rather than sentences, because much punctuation was added by publishers and characters do not always speak in full sentences.
  • Raw relative entropy becomes infinite when a word pair is missing from a profile, so the sum is restricted to transitions that are non-zero in every candidate profile. This also removes a bias toward small profiles.
  • Attribution accuracy across all authors is 92.6%, higher than frequency-based baselines. Later analyses use up to eight profiles: Shakespeare, Marlowe, Chapman, Jonson, Fletcher, Middleton, Peele and Greene. Kyd could not be profiled, since his only secure play is The Spanish Tragedy.
  • Titus Andronicus:
    • Act 1 goes to Peele and the rest to Shakespeare, which supports Weber's case that Shakespeare wrote Scene 4.1.
    • For the "Fly" scene (3.2), Marlowe scores 0.27 and Shakespeare 0.47. Shakespeare is still preferred, because the scene appeared only after Marlowe's death.
    • Middleton (3.72) is not supported as the author of this scene.
  • Edward III: Act 1 goes to Marlowe. Scenes 3.2, 3.3, 4.1, 4.2, 4.3, 4.5 and 4.9 sit far from both the Shakespeare and Marlowe profiles, which hints at a third, unprofiled writer.

Interesting ideas

  • I1. Treating directed, distance-discounted co-occurrence of grammatical words as a Markov chain gives a compact fingerprint of style that ignores topic words, so it carries across subjects.
  • I2. Comparing distributions only on the support shared by all candidates stops large, dense profiles from winning simply because they are bigger.
  • I3. The text unit was matched to the genre (speeches instead of sentences) because the original punctuation could not be trusted.
  • I4. A text that is far from every candidate profile is read as evidence of an outside author, not forced onto the nearest one.
  • I5. The top statistical candidate was overruled by a hard chronological fact, because the scene appeared after Marlowe's death.

For Mova

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

  • I1 (tokens): Function-word transition profiles per genre could flag input whose closed-class patterns differ from the tagger's training data, warning that morphology and tags may be less reliable.
  • I2 (learn): When scoring witness parsers or silver corpora of different sizes against each other, limit the comparison to shared relation types so that sheer coverage does not bias the agreement scores.
  • I3 (tree): For dialogue, plays and fables, speaker turns could serve as the parsing and evaluation unit when sentence boundaries are editorial or unreliable.
  • I4 (learn): The same open-set rule can act as a negative control, where inputs distant from all known profiles are rejected or marked unknown instead of being labelled with confidence.
  • I5 (world): Compiled facts about dates and the order of events could veto statistically preferred readings that are impossible in the world of the text.

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

🎓University of PennsylvaniaPhiladelphia · United States170 works · 60,739 citations🎓Massachusetts Institute of TechnologyCambridge · United States159 works · 65,157 citations🎓De Montfort UniversityLeicester · United Kingdom2 works · 24 citations

📚 Shelves 7 Computer Science · 9,086 works Artificial Intelligence · 8,889 works Arts and Humanities · 9,296 works Language and Linguistics · 9,206 works Decision Sciences · 85 works Literature and Literary Theory · 340 works Statistics, Probability and Uncertainty · 32 works

🧭 Topics 1 English · 1,373 works

Cite

Cite as Mark Eisen, Santiago Segarra, Gabriel Egan and Alejandro Ribeiro. 2016. Stylometric Analysis of Early Modern Period English Plays. arXiv:1610.05670.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.1610.05670 · arXiv 1610.05670

@misc{eisen2016stylometric,
  title = {{Stylometric Analysis of Early Modern Period English Plays}},
  author = {Mark Eisen, Santiago Segarra, Gabriel Egan and Alejandro Ribeiro},
  year = {2016},
  doi = {10.48550/arxiv.1610.05670},
  eprint = {1610.05670},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1610.05670}
}

References

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

  1. Shlomo Argamon and Shlomo Levitan. Measuring the usefulness of function words for authorship attribution. In ACH/ALLC, 2005. find
  2. John Burrows. ’Delta’: A measure of stylistic difference and a guide to likely authorship. Literary and linguistic computing, 17(3):267–287, 2002. ↗
  3. Chadwyck-Healey/ProQuest. Literature Online. URL http://lion.chadwyck.com/. Hugh Craig and Arthur F. Kinney. Shakespeare, computers, and the mystery of authorship. Cambridge University Press, 2009. ↗
  4. Doug Cutting, Julian Kupiec, Jan Pedersen, and Penelope Sibun. A practical part-of-speech tagger. In Proceedings of the third conference on Applied natural language processing, pages 133–140. Association for Computational Linguistics, 1992. our card
  5. Olivier De Vel, Alison Anderson, Malcolm Corney, and George Mohay. Mining e-mail content for author identification forensics. ACM Sigmod Record, 30(4):55–64, 2001. ↗
  6. William Shakespeare (Eds. Gwynne B. Evans and John Joseph M. Tobin). The Riverside Shakespeare, volume 1. Houghton Mifflin Boston, 1974. find
  7. Ed. Alan B. Farmer and Zachary Lesser. DEEP: Database of Early English Playbooks. http://deep.sas.upenn. edu/ , 2007. URL http://deep.sas.upenn.edu/. ↗
  8. Frederick Gard Fleay. Shakespeare Manual. Ams PressInc, 1878. find
  9. Philip Gaskell. A new introduction to bibliography. Clarendon Press Oxford, 1972. find
  10. David I. Holmes. Vocabulary richness and the prophetic voice. Literary and Linguistic Computing, 6(4): 259–268, 1991. ↗
  11. David I. Holmes. A stylometric analysis of mormon scripture and related texts. Journal of the Royal Statistical Society. Series A (Statistics in Society), pages 91–120, 1992. ↗
  12. David L. Hoover. Another perspective on vocabulary richness. Computers and the Humanities, 37(2):151– 178, 2003. ↗
  13. Edwin J. Howard. The printer and Elizabethan punctuation. Studies in Philology, pages 220–229, 1930. find
  14. Cyrus Hoy. The shares of Fletcher and his collaborators in the Beaumont and Fletcher canon (i). Studies in Bibliography, 8:129–146, 1956. find
  15. Cyrus Hoy. The shares of Fletcher and his collaborators in the Beaumont and Fletcher canon (v). Studies in Bibliography, 12:77–108, 1960. find
  16. MacDonald P. Jackson. Defining Shakespeare: ’Pericles’ as Test Case. Oxford University Press, 2003. find
  17. MacDonald P. Jackson. Shakespeare and the quarrel scene in Arden of Faversham. Shakespeare Quarterly, 57(3):249–293, 2006. ↗
  18. Patrick Juola. Authorship attribution. Foundations and Trends in information Retrieval, 1(3):233–334, 2006. find
  19. George Kesidis and Jean Walrand. Relative entropy between markov transition rate matrices. IEEE transactions on information theory, 39(3):1056–1057, 1993. ↗
  20. Dmitri V. Khmelev and Fiona J. Tweedie. Using Markov chains for identification of writer. Literary and linguistic computing, 16(3):299–307, 2001. find
  21. Terence P. Logan and Denzell S. Smith. The New Intellectuals. University of Nebraska Press, 1977. find
  22. Norman Meuschke and Bela Gipp. State-of-the-art in detecting academic plagiarism. International Journal for Educational Integrity, 9(1), 2013. ↗
  23. Frederick Mosteller and David Wallace. Inference and disputed authorship: The federalist. 1964. find
  24. Conrad Sanderson and Simon Guenter. Short text authorship attribution via sequence kernels, Markov chains and author unmasking: An investigation. In Proceedings of the 2006 Conference on Empirical Methods in Natural Language Processing, pages 482–491. Association for Computational Linguistics, 2006. find
  25. Santiago Segarra, Mark Eisen, and Alejandro Ribeiro. Authorship attribution using function words adjacency networks. In Acoustics, Speech and Signal Processing (ICASSP), 2013 IEEE International Conference on, pages 5563–5567. find
  26. Santiago Segarra, Mark Eisen, and Alejandro Ribeiro. Authorship attribution through function word adjacency networks. IEEE Transactions on Signal Processing, 63(20):5464–5478, Oct 2015. ↗
  27. Santiago Segarra, Mark Eisen, Gabriel Egan, and Alejandro Ribeiro. Attributing the authorship of the Henry VI plays by word adjacency. Shakespeare Quarterly, 67(2):232–256, 2016. ↗
  28. Efstathios Stamatatos. A survey of modern authorship attribution methods. Journal of the American Society for information Science and Technology, 60(3):538–556, 2009. ↗
  29. Gary Taylor and Doug Duhaime. Who wrote the Fly Scene (3.2) in Titus Andronicus? Automated searches and deep reading. In Gary Taylor and Gabriel Egan, editors, The New Oxford Shakespeare Authorship Companion, pages 67–91. Oxford University Press, 2017. find
  30. Gary Taylor and John Jowett. Shakespeare Reshaped, 1606-1623. Cambridge Univ Press, 1993. find
  31. Gary Taylor and John Lavagnino. Thomas Middleton: The Collected Works. Oxford University Press, 2007. find
  32. Gary Taylor and Rory Loughnane. The canon and chronology of Shakespeare’s works. In Gary Taylor and Gabriel Egan, editors, The New Oxford Shakespeare Authorship Companion, pages 417–602. Oxford University Press, 2017. find
  33. Richard W. Van Fossen. ’Eastward Ho’ by George Chapman, Ben Jonson, and John Marston. Manchester University Press, 1979. find
  34. Brian Vickers. Shakespeare, Co-Author: A Historical Study of the Five Collaborative Plays. Oxford University Press, 2002. find
  35. William W Weber. Shakespeare after all?: The authorship of titus andronicus 4.1 reconsidered. SHAKE-SPEARE SURVEY, VOL 67: SHAKESPEARE’S COLLABORATIVE WORK, 67:69–84, 2014. find
  36. Archie Webster. Was Marlowe the man? National Review, pages 81–6, 1923. find
  37. Stanley Wells. Shakespeare and Co.: Christopher Marlowe, Thomas Dekker, Ben Jonson, Thomas Middleton, John Fletcher and the Other Players in His Story. Random House LLC, 2009. find
  38. George Udny Yule. On sentence-length as a statistical characteristic of style in prose: With application to two cases of disputed authorship. Biometrika, pages 363–390, 1939. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/eisen-2016-stylometric-analysis-early-modern-period/