Elamur Research⌕
☰
home / papers / 1986 / Discovery Procedures for Sublanguage Selectional Patterns: Initial Experiments

Discovery Procedures for Sublanguage Selectional Patterns: Initial Experiments

Paper · 1986 · Ralph Grishman, Lynette Hirschman and Ngo Thanh Nhan

Summary

Selectional constraints are domain-specific rules about which semantic word classes may combine in clause structures, prepositional phrases, prenominal adjectives and compound nouns. They help a parser reject wrong analyses, but they must be rebuilt for each new subject area, which makes systems hard to port. Grishman, Hirschman and Nhan explore a semi-automated way to acquire these rules. They parse a sample of sublanguage text, in this case medical hospital discharge summaries, with the Linguistic String Parser. They then collect the class co-occurrence patterns found in the analyses and reuse them to constrain parsing of further texts. A relaxation procedure compensates for patterns missing from the collected set. A baseline experiment shows that selection hardly changes how many sentences receive correct parses, but it sharply reduces incorrect ones, so undetected errors become detectable failures. Patterns gathered from a modest sample, combined with relaxation, perform roughly as well as hand-crafted ones. The authors stress that substantial human supervision remains necessary. They suggest combining text-driven collection with interactive elicitation from domain experts.

Key points

  • The test corpus contained 407 sentences and fragments from hospital discharge summaries. It was parsed with the NYU Linguistic String Project medical grammar, an augmented context-free grammar with restrictions written in Restriction Language.
  • With manually built selection, results were 76% good, 7% bad and 17% no parse. Without selection, they were 75% good, 17% bad and 8% no parse.
  • With selection, 21 sentences went from a good parse to no parse because the constraints were too strict. Another 23 moved from a bad parse to a good or acceptable one.
  • The growth curves for discovered patterns indicate that fuller coverage and better parsing would come from a corpus several times the current size.
  • Planned next steps include an interactive environment for resolving homographs and ambiguities and for reviewing newly found patterns. The authors also propose bootstrapping through expert elicitation, citing a BBN system that generalizes patterns via a class hierarchy.

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

🎓New York UniversityNew York · United States201 works · 100,180 citations

 Fields, topics and concepts

📚 Shelves 4The 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 works

Cite

Cite as Ralph Grishman, Lynette Hirschman and Ngo Thanh Nhan. 1986. Discovery Procedures for Sublanguage Selectional Patterns: Initial Experiments. Computational Linguistics.

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

@article{grishman1986discovery,
  title = {{Discovery Procedures for Sublanguage Selectional Patterns: Initial Experiments}},
  author = {Ralph Grishman, Lynette Hirschman and Ngo Thanh Nhan},
  journal = {Computational Linguistics},
  year = {1986},
  url = {https://aclanthology.org/J86-3002/}
}

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. Ballard, B.; Lusth, J.; and Tinkham, N. 1984 LDC-I: A Transportable Knowledge-Based Natural Language Processor for Office Environments. ACM Transactions on Office Information Systems 2: 1-25. find
  2. Chevalier, M. et al. 1978 TAUM-METEO: Description du Syst~me. Groupe de recherche en traduction automatique, Universit6 de Montr6al. find
  3. Grishman, R. and Hirschman, L. 1982 Natural Language Interfaces Using Limited Semantic Information. NSF Final Report, New York University, New York, New York. find
  4. Grishman, R.; Hirschman, L.; and Friedman, C. 1982 Natural Language Interfaces using Limited Semantic Information. Proceedings of the 9th International Conference on Computational Linguistics. Prague, Czechoslovakia: 89-94. find
  5. Grishman, R.; Hirsehman, L.; and Friedman, C. 1983 Isolating Domain Dependencies in Natural Language Interfaces. Proceedings of the Conference on Applied Natural Language Processing. Santa Monica, California: 46-53. find
  6. Grishman, R.; Nhan, N. T.; and Marsh, E. 1984 Tuning Natural Language Grammars for New Domains. Proceedings o f the Conference on Intelligent Systems and Machines. Rochester, Minnesota: 342-346. find
  7. Grishman, R.; Nhan, N. T.; Marsh, E.; and Hirschman, L. 1984 Automated Determination of Sublanguage Syntactic Usage. Proceedings of COLING84 (Tenth International Conference on Computational Linguistics). Stanford, California: 96-100. find
  8. Grosz, B. 1983 TEAM: A Transportable Natural-Language Interface System. Proceedings of Conference on Applied Natural Language Processing. Santa Monica, California: 39-45. find
  9. Harris, Z. 1968 Mathematical Structures of Language. Wiley Interscience, New York, New York. find
  10. Hirschman, L.; Grishman, R.; and Sager, N. 1975 Grammatically-based Automatic Word Class Formation. Information Processing and Management 11 : 39-57. find
  11. Hirschman, L.; Story, G.; Marsh, E.; Lyman, M.; and N. Sager. 1981 An Experiment in Automated Health Care Evaluation of Narrative Medical Records. Computers and Biomedical Research 14: 447-463. find
  12. Hirschman, L. and Sager, N. 1983 Automatic Information Formatting of a Medical Sublanguage. In Kittredge and Lehrberger: 27-80. find
  13. Kittredge, R. and Lehrberger, J., Eds. 1983 Sublanguage: Studies of Language in Restricted Semantic Domains. Series of Foundations of Communications, Walter de Gruyter, Berlin: 27-80. find
  14. Lehrberger, J. 1983 Automatic Translation and the Concept of Sublanguage. In Kittredge and Lehrberger. find
  15. Ralph Grishman, Lynette Hirsehman, Ngo Thanh Nhan Marsh, E. 1983 Utilizing Domain-Specific Information for Processing Compact Text,~ Proceedings of Conference on Applied Natural Language Processing. Santa Monica, California: 99-103. find
  16. Marsh, E.; Hamburger, H.; and Grishman, R. 1984 A Production Rule System for Message Summarization. Proceedings of the 1984 National Conference on Artificial Intelligence. Austin, Texas. find
  17. Sager, N. 1981 Natural Language Information Processing: A Computer Grammar of English and its Applications. Addison-Wesley, Reading, Massachusetts. find
  18. Sager, N. and Grishman, R. 1975 The Restriction Language for Computer Grammars of Natural Language. Communications ACM 18: 390-400. find
  19. Weischedel, R. M. and Sondheimer, N . K . 1983 Meta-rules as a Basis for Processing Ill-Formed Input. Journal of Computational Linguistics 9: 161-177. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/grishman-1986-discovery-procedures-sublanguage-selectional-patterns/