Modeling English Past Tense Intuitions with Minimal Generalization

What it is about

The paper proposes a learner that takes pairs of related forms (present and past) and induces rules by minimal generalisation: each pair gives a word-specific rule with a structural change and a context, and rules with the same change are merged into the most specific context that covers both. Every rule gets a confidence, a lower confidence limit on its reliability (hits over scope, smoothed and adjusted with a parameter alpha), and a novel form takes the output of the most confident applicable rule. A later step, impugnment, penalises a broad context when a narrower subcontext does most of its work. The model is trained on 4,253 CELEX verbs with frequency of at least 10. It gives the regular past for almost every training verb and produces plausible forms for nonce verbs. In the rating study the model's predictions correlate with the mean participant ratings at r = .745 for regulars and r = .570 for irregulars (n = 41), and at r = .695 and .333 for volunteered forms. The authors note that impugnment had to be switched off to match the data best.

Ideas

• I1. Implement the rule learner in Rust as a new module next to en/src/morph.rs: rules as records with scope and hits counted over BTreeMap-ordered forms, confidence from the lower bound formula, and a why string that prints the rule that fired. This is the transparency the project needs, and it is deterministic.
• I2. Impugnment as a measured option, not a default: run the learner with and without the subcontext penalty on the past-tense training list, and report the difference on held-out verbs. The paper itself found the penalty hurt the match to ratings.
• I3. Use the four-category design (regular-only, irregular-only, both, neither) as a test split for any nonce-verb evaluation in Mova, reporting accuracy per category.
• I4. Negative control: shuffle the stems against their past forms; the learner's confidence on regular forms must fall sharply, or the scoring is not using the stem.
• I5. Multi-change and distributional-encroachment cases (for example burnt, dwelt) as a list of known exceptions to test impugnment on: a learner that cannot tell them apart from the regular rule fails the test.

For Mova

The chapters are 14. Irregular verbs and 13. Verbs. The code would sit next to en/src/morph.rs, with en morph-eval as the measure. Report accuracy on held-out verbs, and keep the nonce-word ratings for evaluation only, under their licence. The paper's data are NC-licensed: use them as a summary only and do not carry them into a release.

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.3115/1118647.1118654

@inproceedings{albright2002modeling,
  title = {{Modeling English Past Tense Intuitions with Minimal Generalization}},
  author = {Albright, Adam and Hayes, Bruce},
  booktitle = {Proceedings of the ACL-02 Workshop on Morphological and Phonological Learning},
  year = {2002},
  doi = {10.3115/1118647.1118654},
  url = {https://doi.org/10.3115/1118647.1118654}
}

References

31 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Albright, Adam. 1999. Phonological subregularities in inflectional classes: Evidence from Italian. In Matthew Gordon, ed., UCLA Working Papers in Linguistics, Vol. 1 (Papers in Phonology 2), pp. 1-47. Albright, Adam. To appear. The productivity of infixation in Lakhota. To appear in Pamela Munro, ed., UCLA Working Papers in Linguistics (Studies in Lakhota). find
2. Albright, Adam and Bruce Hayes 1999. Distributional encroachment and its consequences for phonological learning. UCLA Working Papers in Linguistics 4 (Papers in Phonology 4), 179-190. find
3. Albright, Adam and Bruce Hayes. 2001. Rules vs. analogy in English past tenses: A computational/experimental study. Ms., Department of Linguistics, UCLA. http://www.linguistics.ucla.edu/people/hayes/rulesvsanalogy/ ↗
4. Archangeli, Diana and Douglas Pulleyblank. 1987. Maximal and minimal rules: effects of tier scansion. In Joyce McDonough and Bernadette Plunkett, eds., Proceedings of the North Eastern Linguistic Society 17, Graduate Linguistics Student Association, University of Massachusetts, Amherst, pp. 16-35. find
5. Baroni, Marco. 2000. Distributional cues in morpheme discovery: A computational model and empirical evidence. Ph.D. dissertation, UCLA. http://www.ai.univie.ac.at/~marco/ Baroni Marco, Johannes Matiasek, and Trost Harald. 2002. Unsupervised discovery of morphologically related words based on orthographic and semantic similarity. To appear in Proceedings of the Workshop on Morphological and Phonological Learning of ACL-2002. http://www.ai. univie.ac.at/~marco/ Bayardo, R., R. Agrawal and D. Gunopulos. 1999. Constraint-based rule mining in large, dense databases. In ICDE-99. http://citeseer.nj.nec.com/ba-yardo99constraintbased.html ↗ our card
6. Burnage, G. 1991. CELEX - A Guide for Users. Nijmegen: Centre for Lexical Information, University of Nijmegen. find
7. Chomsky, Noam and Morris Halle. 1968. The Sound Pattern of English. New York: Harper and Row. find
8. Daugherty, Kim G. and Mark S. Seidenberg. 1994. Beyond rules and exceptions: a connectionist approach to inflectional morphology. In S. D. Lima, R. L. Corrigan and G. K. Iverson, eds., The Reality of Linguistic Rules. Amsterdam: J. Benjamins. find
9. Dixon, Robert M. W. 1977. A Grammar of Yidin y . Cambridge: Cambridge University Press. find
10. Dzeroski, Saso and Tomaz Erjavec. 1997. Learning Slovene declensions with FOIDL. In W. Daelemans, A. Van den Bosch, and A. Weijters, eds., Workshop Notes of the ECML/Mlnet Workshop on Empirical Learning of Natural Language Processing Tasks, Prague, pp. 49-60. find
11. Eddington, David. 2000. Analogy and the dual-route model of morphology. Lingua 110:281–298. ↗
12. Friederici, A. D. and Wessels, J. E. 1993. Phonotactic knowledge of word boundaries and its use in infant speech perception. Perception and Psychophysics 54:287–295. ↗
13. Goldsmith, John. 2001. Unsupervised learning of the morphology of a natural language. Computational Linguistics 27:153-198. ↗
14. Halle, Morris. 1978. Knowledge unlearned and untaught: what speakers know about the sounds of their language. In Morris Halle, Joan Bresnan, and George Miller, eds., Linguistic Theory and Psychological Reality. Cambridge, MA: MIT Press, pp. 294-303. find
15. Hayes, Bruce. 1982. Metrical structure as the organizing principle of Yidin y phonology. In H. van der Hulst and N. Smith, eds., The Structure of Phonological Representations, Part I. Dordrecht: Foris Publications, pp. 97-110. find
16. Jusczyk, P. W., Friederici, A. D., Wessels, J. M., Svenkerud, V. Y. and Jusczyk, A. M. (1993). Infants’ sensitivity to the sound patterns of native language words. Journal of Memory and Language 32:402– 420. ↗
17. Jusczyk, P. W., Luce, P. A. and Charles-Luce, J. 1994. Infants’ sensitivity to phonotactic patterns in the native language. Journal of Memory and Language 33:630–645. ↗
18. Kruskal, J. B. 1983. An overview of sequence comparison. In D. Sankoff and J. B. Kruskal, eds., Time Warps, String Edits, and Macromolecules: The Theory and Practice of Sequence Comparison. Reading, MA: Addison-Wesley Publishing Company. find
19. Ling, Charles X. and Marin Marinov. 1993. Answering the connectionist challenge: A symbolic model of learning the past tenses of English verbs. Cognition 49:235–290. ↗
20. McCarthy, John. 1979. On stress and syllabification. Linguistic Inquiry 10:443-465. find
21. MacWhinney, Brian and Jared Leinbach. 1991. Implementations are not conceptualizations: Revising the verb learning model. Cognition 40:121-157. ↗
22. Mester, Armin. 1994. The quantitative trochee in Latin. Natural Language and Linguistic Theory 12: 1-61. ↗
23. Mikheev, Andrei. 1997. Automatic rule induction for unknown-word guessing. Computational Linguistics 23:405–423. find
24. Mooney, Raymond J. and Mary Elaine Califf. 1995. Learning the past tense of English verbs using inductive logic programming. In Stefan Wermter, Ellen Riloff, and Gabriele Scheler, eds., Connectionist, Statistical, and Symbolic Approaches to Learning for Natural Language Processing. Springer Verlag, pp. 370-384 Nakisa, Ramin. find
25. C., Kim Plunkett, and Ulrike Hahn. 2001. A cross-linguistic comparison of single and dual-route models of inflectional morphology. In P. Broeder and J. Murre, eds., Models of Language Acquisition: Inductive and Deductive Approaches. Cambridge, MA: MIT Press. Neuvel, Sylvain. To appear. Whole word morphologizer. To appear in Brain and Language. find
26. Pinker, Steven. 1999. Words and Rules: The Ingredients of Language. New York: Basic Books. find
27. Pinker, Steven and Alan S. Prince. 1988. On language and connectionism: Analysis of a parallel distributed processing model of language acquisition. Cognition 28:73–193. ↗
28. Prasada, Sandeep and Pinker, Steven. 1993. Generalization of regular and irregular morphological patterns. Language and Cognitive Processes 8:1–56. Rumelhart, David. find
29. E. and McClelland, Jay L. 1986. On learning the past tenses of English verbs. In D. E. Rumelhart, J. L. McClelland, and The PDP Research Group, eds., Parallel Distributed Processing: Explorations in the Microstructure of Cognition: Vol. 2. Cambridge, MA: MIT Press, pp. 216-271. find
30. Schone, Patrick and Daniel Jurafsky. 2001. Knowledge-free induction of inflectional morphologies. Proceedings of the North American chapter of the Association for Computational Linguistics (NAACL-2001). http://www.colorado.edu/linguistics/jurafsky/pubs. html. ↗ our card
31. Simon, Anthony and Alan Frisch. 2001. Cautious induction in inductive logic programming. In ILP97, The Seventh International Workshop on Inductive Logic Programming, pp. 45-60. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/albright-2002-modeling-english-past-tense-intuitions/