Elamur Research⌕
☰
home / papers / 2006 / Extending VerbNet with Novel Verb Classes

Extending VerbNet with Novel Verb Classes

Paper · 2006 · Karin Kipper, Anna Korhonen, Neville Ryant et al.

Summary

Levin's 1993 verb classification, and VerbNet (VN) built on top of it, mostly describe verbs whose arguments are noun phrases and prepositional phrases. Frequent verbs that take adjectival, adverbial, particle, predicative, control or sentential complements were left largely uncovered. Korhonen and Briscoe (K&B, 2004) had proposed 57 new Levin-style classes for such verbs. However, those classes had no detailed syntactic-semantic descriptions and were not organised into any taxonomy.

This paper merges the K&B classes into VerbNet. Each class receives VN-style thematic roles, selectional restrictions, syntactic frames and semantic predicates. Where a new class overlapped with an existing one, VN was reorganised: some classes became subclasses, others were merged, and some old classes were revised. The extended lexicon grows from 191 to 237 first-level classes and gains many new semantic predicates and sentential-complement restrictions. The authors describe the result as the broadest Levin-style English verb classification, freely available. It is also meant to serve as a stronger gold standard for future automatic class induction. The extended VN itself was not evaluated on downstream tasks.

Key points

  • VN 1.0 before the extension:

    • 4,100 verbs (about 3,000 lemmas) in 191 first-level classes plus 74 subclasses;
    • 21 thematic roles and 64 semantic predicates, together with 36 selectional restrictions on arguments and 314 syntactic frames;
    • its frames account for over 78% of exact matches on almost 50,000 PropBank instances.

  • How K&B built their classes, semi-automatically:

    • they first derived 106 new diathesis alternations from pairs of the 163 SCFs in Briscoe (2000);
    • they then screened 102 candidate classes drawn from Rudanko, Sager, Levin and the LCS database;
    • they finally kept 57 classes of 2–45 verbs each, adding members found through WordNet relations.

  • Example K&B classes include ORDER, URGE, FORCE, WISH, ALLOW, FORBID, HELP, DEDICATE and LECTURE. ORDER, for instance, links SCFs 57, 104 and 106 through alternations.

  • 13 proposed classes overlapped heavily with existing VN classes and required restructuring. For example, WANT overlapped with Want-32.1, PAY with Give-13.1, and SEE with See-30.1. Other proposed classes were added as subclasses, such as Turn-26.6.1, Convert-26.6.2 and Shift-26.6.3.

  • Extended VN compared with VN 1.0:

    MeasureVN 1.0Extended VN
    First-level classes191237
    Thematic roles2123
    Semantic predicates6494
    Sentential-complement restrictions355
    Lemmas3,0073,175
    Verb senses4,1734,526

Interesting ideas

  • I1. A class is accepted only if its members' subcategorization frames can be linked by attested diathesis alternations, which gives an objective test for whether a meaning grouping is also syntactically coherent.
  • I2. Thin candidate classes are padded with WordNet synonyms, troponyms, hypernyms, coordinate terms and antonyms of existing members, then checked against syntax dictionaries.
  • I3. Overlapping classes are resolved by taking the union of members and frames when meaning and syntax cannot separate them, and by re-cutting old heterogeneous classes when a new class exposes the problem (as CONSIDER did for four predicative classes).
  • I4. Finite and nonfinite sentential complements are told apart by syntactic restrictions on the role (as in Tell-37.2), so one role can map to different clause types.
  • I5. Each semantic predicate is tied to start, during or end of the event, which encodes aspect and the resulting state directly in the frame.

For Mova

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

  • I1 (world): When compiling verb-class knowledge, we could keep only groupings whose frames are linked by alternations, and reject or split the rest.
  • I2 (learn): Proposed verb-class members could be generated from WordNet neighbours and accepted only when silver parses show the expected frames.
  • I3 (world): Our class inventory could carry explicit merge and split rules, so that adding a resource triggers cleanup of incoherent classes instead of producing duplicates.
  • I4 (frames): Role assignment for "tell X that…" versus "tell X to…" could use these restrictions to choose a proposition-type or action-type Topic.
  • I5 (world): These event-stage markers could drive state updates and event ordering in the text world, for example by asserting a result only at end(E).

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 States221 works · 63,234 citations🎓University of CambridgeCambridge · United Kingdom243 works · 121,396 citations🎓University of Colorado BoulderBoulder · United States69 works · 51,069 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksPsychology · 1,552 worksExperimental and Cognitive Psychology · 870 works

🧭 Topics 1Research subjects and languages the work is about.Semantic role labeling · 145 works

Cite

Cite as Karin Kipper, Anna Korhonen, Neville Ryant and Martha Palmer. 2006. Extending VerbNet with Novel Verb Classes. Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC'06).

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

@inproceedings{kipper2006extending,
  title = {{Extending VerbNet with Novel Verb Classes}},
  author = {Karin Kipper, Anna Korhonen, Neville Ryant and Martha Palmer},
  booktitle = {Proceedings of the Fifth International Conference on Language Resources and Evaluation (LREC'06)},
  year = {2006},
  doi = {10.63317/2z6m7csxuxbs},
  url = {https://doi.org/10.63317/2z6m7csxuxbs}
}

References

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

  1. Collin F. Baker, Charles J. Fillmore, and John B. Lowe. 1998. The Berkeley FrameNet project. In Proceedings of the 17th International Conference on Computational Linguistics (COLING/ACL-98), pages 86–90, Montreal. ↗
  2. Branimir Boguraev, Ted Briscoe, John Carroll, David Carter, and Claire Grover. 1987. The derivation of a grammatically-indexed lexicon from the Longman Dictionary of Contemporary English. In Proceedings of the 25 th annual meeting of ACL, pages 193–200, Stanford, CA. Chris Brew and Sabine Schulte im Walde. 2002. Spectral clustering for German verbs. In Conference on Empirical Methods in Natural Language Processing, Philadelphia, USA.
  3. Ted Briscoe. 2000. Dictionary and System Subcategorisation Code Mappings. Unpublished manuscript, http://www.cl.cam.ac.uk/users/alk23/subcat/sub-cat.html, University of Cambridge Computer Laboratory. ↗
  4. Dick Crouch and Tracy Holloway King. 2005. Unifying lexical resources. In In Proceedings of Interdisciplinary Workshop on the Identification and Representa-1031 tion of Verb Features and Verb Classes, Saarbruecken, Germany. find
  5. Hoa Trang Dang. 2004. Investigations into the Role of Marc Moens and Mark Steedman. 1988. Temporal Ontol-Lexical Semantics in Word Sense Disambiguation. Ph.D. ogy and Temporal Reference. Computational Linguis-thesis, CIS, University of Pennsylvania. tics, 14:15–38. find
  6. Detlef Prescher, Stefan Riezler, and Mats Rooth. 2000. find
  7. Bonnie J. Dorr. 1997. Large-scale dictionary construction Using a probabilistic class-based lexicon for lexical for foreign language tutoring and interlingual machine ambiguity resolution. In 18th International Confer-translation. Machine Translation, 12(4):271–325. ence on Computational Linguistics, pages 649–655, Bonnie J. Dorr. 2001. LCS Verb Database. In Online Saarbrücken, Germany. Software Database of Lexical Conceptual Structures and Juhani Rudanko. 1996. Prepositions and Complement Documentation, University of Maryland. Clauses. State University of New York Press, Albany. find
  8. Christiane Fellbaum, editor. 1998. WordNet: An Eletronic Juhani Rudanko. 2000. Corpora and Complementation. Lexical Database. Language, Speech and Communica-University Press of America. tions. MIT Press, Cambridge, Massachusetts. find
  9. Naomi Sager. 1981. Natural Language Information Pro-Ralph Grishman, Catherine Macleod, and Adam Meyers. cessing: A Computer Grammar of English and Its Appli-1994. COMLEX syntax: building a computational lexi-cations. Addison-Wesley Publising Company, MA. con. In Proceedings of the International Conference on Lei Shi and Rada Mihalcea. 2005. Putting pieces together: Computational Linguistics, Kyoto, Japan. Combining FrameNet, VerbNet and WordNet for robust Svetlana Hensman and John Dunnion. 2004. Automat-semantic parsing. In Proceedings of the Sixth Inter-ically building conceptual graphs using VerbNet and national Conference on Intelligent Text Processing and WordNet. In In Proceedings of the 3rd International Computational Linguistics, Mexico City, Mexico. Symposium on Information and Communication Tech-Robert Swier and Suzanne Stevenson. 2004. Unsupervised nologies (ISICT), pages 115–120, Las Vegas, NV. semantic role labelling. In Proceedings of the 2004 Con-Ray Jackendoff. 1990. Semantic Structures. MIT Press, ference on Empirical Methods in Natural Language Pro-Cambridge, Massachusetts. cessing, pages 95–102, Barcelona, Spain, August. find
  10. Paul Kingsbury and Martha Palmer. 2002. find
  11. From Treebank Mary Swift. 2005. Towards automatic verb acquisition to PropBank. In Proceedings of the 3rd International from VerbNet for spoken dialog processing. In In Pro-Conference on Language Resources and Evaluation, Las ceedings of Interdisciplinary Workshop on the Identifi-Palmas, Canary Islands, Spain. cation and Representation of Verb Features and Verb Paul Kingsbury. 2004. Verb clusters from PropBank an-Classes, Saarbruecken, Germany. notation. Technical report, University of Pennsylvania, XTAG Research Group. 2001. A lexicalized tree adjoin-Philadelphia, PA. ing grammar for English. Technical Report IRCS-01-03, IRCS, University of Pennsylvania. find
  12. Karin Kipper, Hoa Trang Dang, and Martha Palmer. 2000. Class-based construction of a verb lexicon. In AAAI/IAAI, pages 691–696. find
  13. Karin Kipper-Schuler. 2005. VerbNet: A broad-coverage, comprehensive verb lexicon. Ph.D. thesis, Computer and Information Science Dept., University of Pennsylvania, Philadelphia, PA, June. ↗
  14. Anna Korhonen and Ted Briscoe. 2004. Extended Lexical-Semantic Classification of English Verbs . In Proceedings of the HLT/NAACL Workshop on Computational Lexical Semantics, Boston, MA. find
  15. Anna Korhonen, Yuval Krymolowski, and Zvika Marx. 2003. Clustering polysemic subcategorization frame distributions semantically. In Proceedings of the 41st Annual Meeting of ACL, pages 64–71, Sapporo, Japan. find
  16. Anna Korhonen. 2002. Semantically motivated subcategorization acquisition. In ACL Workshop on Unsupervised Lexical Acquisition, Philadelphia. find
  17. Beth Levin. 1993. English Verb Classes and Alternation, A Preliminary Investigation. The University of Chicago Press. find
  18. Diana McCarthy. 2001. Lexical Acquisition at the Syntax-Semantics Interface: Diathesis Alternations, Subcategorization Frames and Selectional Preferences. Ph.D. thesis, University of Sussex. find
  19. George A. Miller. 1990. WordNet: An on-line lexical database. International Journal of Lexicography, 1032 3(4):235–312. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/kipper-2006-extending-verbnet-novel-verb-classes/