Elamur Research⌕
☰
home / papers / 2018 / Semantic Supersenses for English Possessives

Semantic Supersenses for English Possessives

Paper · 2018 · Austin Blodgett and Nathan Schneider

Summary

Blodgett and Schneider ask whether a semantic label set built for prepositions and postpositions can also describe English possessives. They cover three forms: the 's clitic, possessive pronouns, and the of-genitive. Earlier work used special inventories for noun–noun relations or for the s-genitive only. This paper instead reuses the adposition supersense hierarchy of Schneider et al. (2017), which has 50 categories, almost unchanged.

Two linguists labeled every possessive construction in a 55,000-word corpus of English web reviews. They also set policies for idioms and fixed names. Agreement was checked on a random double-annotated sample.

The main findings:

  • 28 of the 50 supersenses are enough to cover nearly all tokens.
  • The of-genitive and the s-genitive behave very differently. Some relations occur with only one of the two forms.
  • Compared with earlier schemes, the inventory is somewhat finer-grained and keeps the direction of a relation explicit.

The work matters because one cross-linguistic label set now covers both prepositions and possessives. The data is released in STREUSLE v4.0. It supports research on the genitive alternation and on automatic disambiguation.

Key points

  • The corpus contains 1,613 possessive tokens labeled with 28 attested supersenses. Each token gets a single label. 11 idiomatic pronoun uses received a special `$ label.

  • Inter-annotator agreement was measured on 100 sentences containing 109 s-genitive tokens. Raw agreement was 72% and Cohen's κ was 0.66.

  • Labels group into three branches, with clear contrasts between the two forms:

    BranchShare of of-tokensShare of s-genitive tokens
    CONFIGURATION (static relations)83.3%70.3%
    PARTICIPANT roles13.9%29.2%
    CIRCUMSTANCE2.8%0.5%

  • Some labels appear with only one form:

    • The corpus has no clear of-POSSESSOR; ownership is expressed with 's or pronouns.
    • QUANTITY, SPECIES, STUFF, TOPIC, IDENTITY and CHARACTERISTIC occur only with of.
    • BENEFICIARY, ORGROLE and RECIPIENT occur only with the genitive case.

  • The scheme has 29 labels including the idiom category. Tratz & Hovy have 18 labels and Badulescu & Moldovan have 22. The supersenses split pairs by direction, e.g. POSSESSOR vs POSSESSION and GESTALT vs CHARACTERISTIC. The data ships in STREUSLE v4.0.

Interesting ideas

  • I1. One relation inventory serves prepositions, the 's clitic and possessive pronouns alike, so different surface forms of the same relation receive the same meaning label instead of separate tag sets.
  • I2. The labels state which side holds the relation: owner vs owned, property-holder vs property. This direction is what downstream inference needs to decide who has what.
  • I3. Fixed names and possessed idioms are kept out of compositional labeling. Examples are "Ben's Chili Bowl" and "try one's best", where the pronoun must agree with the subject. They are marked as idiom or noncompositional rather than forced into a relation.
  • I4. The two forms have strongly skewed relation distributions: ownership and social ties favor 's, while quantity and material favor of. These skews work as cheap, explainable priors for disambiguation.
  • I5. Agreement was measured on a random sample annotated independently after a calibration discussion. The second annotator skipped tokens they remembered from that discussion, so the agreement figure would not be inflated.

For Mova

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

  • I1 (frames): Map nmod:poss, possessive pronouns and of-nmod edges onto the same role vocabulary already used for prepositional frames, so "the car's owner" and "the owner of the car" yield identical frames.
  • I2 (world): When asserting holdings or attributes from possessives, store the direction explicitly (POSSESSOR→thing vs thing→POSSESSION) so "her dog" and "the owner of a new car" create correctly oriented ownership facts.
  • I3 (world): Detect possessed-idiom patterns before entity building so "I tried my best" does not create a spurious possession of a "best" object.
  • I4 (frames): Use the form (of vs 's/PRP$) as a feature or prior when choosing a role for a possessive, e.g. down-weight POSSESSOR for of-phrases and QUANTITY for 's.
  • I5 (learn): When scoring LLM-teacher or witness-parser agreement, exclude examples used in prompt design or rule tuning so agreement estimates stay honest.

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 WashingtonSeattle · United States263 works · 180,581 citations🎓Georgetown UniversityWashington · United States69 works · 4,417 citations

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 9,086 worksArtificial Intelligence · 8,889 worksArts and Humanities · 9,296 worksLanguage and Linguistics · 9,206 worksPsychology · 798 worksExperimental and Cognitive Psychology · 463 works

🧭 Topics 1Research subjects and languages the work is about.English · 1,373 works

Cite

Cite as Austin Blodgett and Nathan Schneider. 2018. Semantic Supersenses for English Possessives. Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018).

The paper itself, not this page. BibTeX from the library's record of the work.

@inproceedings{blodgett2018semantic,
  title = {{Semantic Supersenses for English Possessives}},
  author = {Austin Blodgett and Nathan Schneider},
  booktitle = {Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)},
  year = {2018},
  doi = {10.63317/2drsyqxd93ry},
  url = {https://doi.org/10.63317/2drsyqxd93ry}
}

References

21 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Badulescu, Adriana and Moldovan, Dan (2009). A Semantic Scattering model for the automatic interpretation of English genitives. Natural Language Engineering, 15(2):215–239. ↗
  2. Baldwin, Timothy and Kim, Su Nam (2010). Multiword expressions. In Indurkhya, Nitin and Damerau, Fred J., editors, Handbook of Natural Language Processing, Second Edition, pages 267–292. CRC Press, Taylor and Francis Group, Boca Raton, FL. find
  3. Bond, Francis, Ho, Jia Qian, and Flickinger, Daniel (2015). Feeling our way to an analysis of English possessed idioms. In Proc. of HPSG, pages 61–74. Singapore. find
  4. Bond, Francis, Sameha, Sheefa Samara, and Flickinger, Dan (2013). Making English possessed idioms our own. In Proc. of HPSG. Berlin, Germany. find
  5. Heine, Bernd (2006). Possession: Cognitive Sources, Forces, and Grammaticalization. Cambridge University Press, Cambridge, UK. find
  6. Hwang, Jena D., Bhatia, Archna, Han, Na-Rae, O’Gorman, Tim, Srikumar, Vivek, and Schneider, Nathan (2017). Double trouble: the problem of construal in semantic annotation of adpositions. In Proc. of *SEM, pages 178– 188. Vancouver, Canada. ↗
  7. Moldovan, Dan, Badulescu, Adriana, Tatu, Marta, Antohe, Daniel, and Girju, Roxana (2004). Models for the semantic classification of noun phrases. In Proceedings of the HLT-NAACL Workshop on Computational Lexical Semantics, pages 60–67. Association for Computational Linguistics. find
  8. Nikiforidou, Kiki (1991). The meanings of the genitive: a case study in semantic structure and semantic change. Cognitive Linguistics, 2(2):149–205. ↗
  9. Quirk, Randolph and Greenbaum, Sidney (1973). A concise grammar of contemporary English. Harcourt School. find
  10. Rosenbach, Anette (2002). Genitive Variation in English: Conceptual Factors in Synchronic and Diachronic Studies. Walter de Gruyter. find
  11. Schneider, Nathan, Hwang, Jena D., Bhatia, Archna, Han, Na-Rae, Srikumar, Vivek, O’Gorman, Tim, and Abend, Omri (2017). Adposition Supersenses v2. arXiv preprint: http://arxiv.org/abs/1704.02134v1 . ↗
  12. Schneider, Nathan, Hwang, Jena D., Bhatia, Archna, Han, Na-Rae, Srikumar, Vivek, O’Gorman, Tim, Moeller, Sarah R., Abend, Omri, Blodgett, Austin, and Prange, Jakob (2018). Adposition and Case Supersenses v2: Guidelines for English. arXiv preprint: https://arxiv.org/abs/1704.02134 . ↗
  13. Schneider, Nathan, Hwang, Jena D., Srikumar, Vivek, Green, Meredith, Suresh, Abhijit, Conger, Kathryn, O’Gorman, Tim, and Palmer, Martha (2016). A corpus of preposition supersenses. In Proc. of LAW X – the 10th Linguistic Annotation Workshop, pages 99–109. Berlin, Germany.
  14. Schneider, Nathan, Onuffer, Spencer, Kazour, Nora, Danchik, Emily, Mordowanec, Michael T., Conrad, Henrietta, and Smith, Noah A. (2014). Comprehensive annotation of multiword expressions in a social web corpus. In Calzolari, Nicoletta, Choukri, Khalid, Declerck, Thierry, Loftsson, Hrafn, Maegaard, Bente, Mariani, Joseph, Moreno, Asuncion, Odijk, Jan, and Piperidis, Stelios, editors, Proc. of LREC, pages 455–461. Reykjavík, Iceland.
  15. Schneider, Nathan, Srikumar, Vivek, Hwang, Jena D., and Palmer, Martha (2015). A hierarchy with, of, and for preposition supersenses. In Proc. of The 9th Linguistic Annotation Workshop, pages 112–123. Denver, Colorado, USA.
  16. Shih, Stephanie, Grafmiller, Jason, Futrell, Richard, and Bresnan, Joan (2012). Rhythm’s role in genitive construction choice in spoken English. find
  17. Stefanowitsch, Anatol (2003). Constructional semantics as a limit to grammatical alternation: The two genitives of english. Topics in English Linguistics, 43:413–444. ↗
  18. Taylor, John R. (1996). Possessives in English: An Exploration in Cognitive Grammar. Clarendon Press, Oxford, UK. find
  19. Tratz, Stephen and Hovy, Eduard (2013). Automatic interpretation of the English possessive. In Proc. of ACL, pages 372–381. Sofia, Bulgaria. find
  20. Viera, Anthony J., Garrett, Joanne M., et al. (2005). Understanding interobserver agreement: the kappa statistic. Fam Med, 37(5):360–363. find
  21. Wolk, Christoph, Bresnan, Joan, Rosenbach, Anette, and Szmrecsanyi, Benedikt (2013). Dative and genitive variability in Late Modern English: Exploring cross-constructional variation and change. Diachronica, 30(3):382–419. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/blodgett-2018-semantic-supersenses-english-possessives/