Elamur Research⌕
☰
home / papers / 2018 / Investigating English Affixes and their Productivity with Princeton WordNet

Investigating English Affixes and their Productivity with Princeton WordNet

Paper · 2018 · Verginica Mititelu

Summary

The paper asks whether the way word senses are shared between a base word and its derivative can reveal how productive an English affix is. The study uses Princeton WordNet 3.0 as a sense-level lexicon instead of a corpus. The author collected base–derived pairs connected by the derivat and derived from relations, plus a few participle links. Duplicates, spelling variants and false pairs were removed. Affixes were then labelled automatically against a list of affixes, with a linguist resolving the cases the list could not handle. Each pair was assigned one of four derivation types (R, D, RD, I) according to how the linked senses of the two words overlap. The types were defined as patterns of set intersection. The author then compared affix behaviour across these types with corpus-based productivity figures from Hay and Baayen (2002), which draw on CELEX. Affixes whose derivatives develop meanings of their own, independent of the base (types D and I), largely coincide with the affixes that produce many hapaxes in corpora, such as -er, -y, -ly and -ness. A sense-structured lexicon can therefore serve as a rough stand-in for corpus measures of productivity.

Key points

  • PWN 3.0 contains 155,287 lemmas and 206,941 word–sense pairs.
  • The pipeline narrowed the data in stages:
    • 77,939 linked pairs were extracted.
    • Removing duplicates left 40,632 pairs; 73 participle pairs were then added.
    • Removing 81 -ise/-ize and 306 -ou-/-o- variants left 40,318 pairs.
    • After manual cleaning, 30,018 annotated pairs remained.
  • Automatic affix detection used a list of 26 prefixes and 54 suffixes; parasynthetic and successive derivations were handled manually.
  • The most frequent affixes were -ness (3,730), -er (3,100), -ly (2,953), -ion (2,469) and -ing (2,102). Every affix in the top ten occurs in all four types:
    • RD dominates for most of them.
    • -ing, -ly and -er lean towards type R.
    • -ion leads in type D.
  • Some affixes are missing from the data because PWN lacks the links: -less never appears and -like appears only once.

Interesting ideas

  • I1. Derivation is treated as a link between individual senses rather than between whole words, so a derivative can share only some of its base's meanings.
  • I2. The four derivation types are defined by the shape of the overlap between linked senses: all senses of one side linked, all of the other side, both, or neither.
  • I3. Affixes are first found by string difference checked against a closed affix list, and a human handles only the leftovers.
  • I4. The pipeline explicitly drops pairs that are only siblings from a common root (skepticism vs. skeptical) or are etymological borrowings.
  • I5. Productivity measured from a lexicon is checked against corpus hapax counts, which validates one proxy with an independent source.

For Mova

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

  • I1 (world): WordNet-derived knowledge could record derivational links per sense, so that "buzzer" inherits only the sound and calling senses of "buzz" when the ontology is compiled.
  • I2 (world): the same set-overlap labels could flag derived words whose meaning has drifted from the base, where inferring a role or meaning from the base would be risky.
  • I3 (tokens): crate en could split derived lemmas into base plus affix using a small affix inventory and send unmatched cases to a review queue instead of guessing.
  • I4 (learn): such pairs make good negative controls when evaluating an induced derivation or lemma-linking rule, since they look derived but are not.
  • I5 (learn): any lexicon-based signal Mova derives could be cross-checked against corpus statistics from silver data before it is trusted.

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

🔬Romanian AcademyBucharest · Romania9 works · 841 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

🧭 Topics 1Research subjects and languages the work is about.English · 2,379 works

Cite

Cite as Verginica Mititelu. 2018. Investigating English Affixes and their Productivity with Princeton WordNet. Proceedings of the 9th Global Wordnet Conference.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2018.gwc-1.7

@inproceedings{mititelu2018investigating,
  title = {{Investigating English Affixes and their Productivity with Princeton WordNet}},
  author = {Verginica Mititelu},
  booktitle = {Proceedings of the 9th Global Wordnet Conference},
  year = {2018},
  doi = {10.18653/v1/2018.gwc-1.7},
  url = {https://doi.org/10.18653/v1/2018.gwc-1.7}
}

References

19 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Mark Aronoff. 1976. Word Formation in Generative Grammar. Cambridge, MA, London, England: MIT Press. find
  2. Harald Baayen. 1992. Quantitative aspects of morphological productivity. In G. E. Booij and J. van Marle (eds.). Yearbook of Morphology 1991. Dordrecht: Kluwer Academic Publishers:109–149. find
  3. R.H. Baayen, R. Piepenbrock, and L. Gulikens. 1995. The CELEX lexical database (release 2) cd-rom. Philadelphia, PA: Linguistic Data Consortium, Univesity of Pennsylvania. find
  4. Verginica Barbu Mititelu. 2012. Adding morphosemantic relations to the Romanian Wordnet. In Proceedings of the Eight International Conference on Language Resources and Evaluation (LREC 2012):2596–2601. find
  5. Verginica Barbu Mititelu. 2013. Increasing the Effectiveness of the Romanian Wordnet in NLP Applications. Computer Science Journal of Moldova, vol. 21, no. 3(63):320-331. find
  6. Laurie Bauer, Rochelle Lieber, and Ingo Plag. 2013. The Oxford Reference Guide to English Morphology. Oxford University Press. find
  7. Ernst Burgschmidt. 1977. Strukturierung, Norm und Productivität in der Wortbildung. In H. E. Brekle and D. Kastovsky (eds.). Perspektiven der Wortbil-dungforschung. Bonn: Bouvier Verlag. find
  8. Tsvetana Dimitrova, Ekaterina Tarpomanova, and Borislav Rizov. 2014. Coping with derivation in the Bulgarian WordNet. In Proceedings of the Seventh Global WordNet Conference (GWC 2014):109–117. find
  9. Christiane Fellbaum (ed.). 1998. WordNet: An Electronic Lexical Database. Cambridge, MA: MIT Press. find
  10. Christiane Fellbaum, Anne Osherson, and Peter E. Clark. 2009. Putting semantics into WordNets “morphosemantic” links. In Proceedings of the Third Language and Technology Conference, Poznan, Poland. [Reprinted in: Responding to Information Society Challenges: New Advances in Human Language Technology. Springer Lecture Notes in Informatics, volume 5603:350-358. find
  11. Jeniffer Hay and Harald Baayen. 2002. Parsing and Productivity. In G. E. Booij and J. van Marle (eds.). Yearbook of Morphology. Dordrecht: Kluwer Academic Publishers:203–235. find
  12. Jennifer Hay and Harald Baayen. 2003. Phonotactics, Parsing and Productivity. Italian Journal of Linguistics, 1:99–130. find
  13. Svetla Koeva. 2008. Derivational and morphosemantic relations in Bulgarian Wordnet. Intelligent Information Systems, 359–368. find
  14. Svetla Koeva, Svetlozara Leseva, Ivelina Stoyanova, Tsvetana Dimitrova, and Maria Todorova. 2016. Automatic Prediction of Morphosemantic Relations. In Proceedings of the Eighth Global WordNet Conference (GWC 2016):168–176. find
  15. Karel Pala and Dana Hlaváčková. 2007. Derivational relations in Czech WordNet. In Proceedings of the Workshop on Balto-Slavonic Natural Language Processing:75–81. find
  16. Maciej Piasecki, Radosław Ramocki, and Marek Maziarz. 2012. Recognition of Polish Derivational Relations Based on Supervised Learning Scheme. In Proceedings of the Eight International Conference on Language Resources and Evaluation (LREC 2012):916–922. find
  17. Ingo Plag. 1999. Morphological Productivity: Structural Constraints in English Derivation. Berlin: De Gruyter. find
  18. Krešimir Šojat, Matea Srebačić, and Marko Tadić. 2012. Derivational and Semantic Relations of Croatian Verbs. Journal of Language Modelling, Vol 0, No 1:111–142. find
  19. Jaan van Marle. 1985. On the Paradigmatic Dimensions of Morphological Creativity. Dordrecht: Foris. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/mititelu-2018-investigating-english-affixes-their-productivity/