Patrick J. Burns

Associate Research Scholar, Digital Projects @ Institute for the Study of the Ancient World / NYU | Formerly Culture Cognition, and Coevolution Lab (Harvard) & Quantitative Criticism Lab (UT-Austin) | Fordham PhD, Classics | LatinCy developer

Sequitur, Non Sequitur: Word Prediction in the Latin Classroom

Abstract for American Classical League Institute 2021

Abstract

Word prediction is ubiquitous technology—for example, as we type on our phones, they suggest what comes _____. (If you filled this blank with ‘next’, you understand the expectations created by language use and fulfilled by word prediction.) With Latin well served by large digital text collections, we can develop computational models for word prediction and related tasks like autocomplete. Yet pedagogical applications for these models are at present limited. This paper first gives a high-level overview of the concepts behind word prediction (e.g. n-grams, Markov chains, word embeddings) and discusses how these align with (non-computational) “prediction drills” used in teaching Latin (cf. Russell 2018, “Read Like a Roman” JCT 19). The second part introduces an assignment for a graduate prose composition seminar that uses language models both to generate Latin sentences and to evaluate the likelihood of student-written sentences, and discusses how to adapt this assignment for other levels.

Works Cited

• Bamman, D., and Burns, P.J. 2020. “Latin BERT: A Contextual Language Model for Classical Philology.” ArXiv:2009.10053 [Cs] (September 21). http://arxiv.org/abs/2009.10053.
• Bengfort, B., Bilbro, R., and Ojeda, T. 2018. Applied Text Analysis with Python: Enabling Language-Aware Data Products with Machine Learning. Sebastopol, CA: O’Reilly.
• Jurafsky, D., and Martin, J.H. 2020. “Speech and Language Processing (3rd Edition, Draft).” December 30. https://web.stanford.edu/~jurafsky/slp3/.
• Markus, D.D., and Ross, D.P. 2004. “Reading Proficiency in Latin through Expectations and Visualization.” Classical World 98 (1): 79–93. doi:10.2307/4352905.
• Russell, K. 2018. “Read Like a Roman: Teaching Students to Read in Latin Word Order.” Journal of Classics Teaching 19 (37): 17–29. doi:10.1017/S205863101800003X.
• Sprugnoli, R., Passarotti, M., and Moretti, G. 2019. “Vir Is to Moderatus as Mulier Is to Intemperans Lemma Embeddings for Latin.” In Proceedings of the Sixth Italian Conference on Computational Linguistics. Bari, Italy. http://ceur-ws.org/Vol-2481/paper69.pdf.