Elamur Research⌕
☰
home / papers / 1997 / Automatic Extraction of Subcategorization from Corpora

Automatic Extraction of Subcategorization from Corpora

Summary

Lexicalist parsers need a subcategorization lexicon that is both accurate and broad. Hand-built resources such as ANLT and COMLEX Syntax contain omissions and errors, lack frequency information, and are costly to adapt to new domains. Briscoe and Carroll present a pipeline that learns verb subcategorization entries, with relative frequencies, directly from corpus text. Sentences are PoS-tagged, lemmatized and parsed by a probabilistic LR parser using a shallow unification grammar. For each verb, an extractor collects argument patterns from the ranked parses. A classifier maps these onto 160 classes, a superset of ANLT and COMLEX Syntax that also covers control and alternations. An evaluator then filters out unreliable hypotheses statistically. On 14 verbs with many complementation patterns, the acquired class rankings were on average 81.4% correct. This is comparable to earlier systems, which handled at most sixteen classes. Feeding the learned frequencies back into the parser improved its crossing-bracket score by about 7%, although the gain was not significant on the small test set. The work shows that corpus-based lexicon acquisition can scale to a fine-grained class inventory and benefit parsing.

Key points

  • The system has six stages: an HMM tagger, a lemmatizer based on the GATE stemmer, a probabilistic LR parser trained on a treebank, a patternset extractor, a pattern classifier and a patternsets evaluator.
  • The grammar has 455 phrase structure rule schemata, with 29 VSUBCAT and 10 PSUBCAT values. It finds at least one parse for 79% of the Susanne corpus (138K words) and runs at about 50 words per CPU second.
  • The 160 classes merge those of ANLT and COMLEX Syntax, plus about 30 added after inspecting corpus patterns that could not be classified. The classifier rejects roughly 15% of extracted patterns as unclassifiable.
  • Acquisition used 1.2 million words of text. Ranking accuracy averaged 81.4% and token recall was 80.9%. Manning's system reached 82% recall on 4.1 million words with a much smaller class set.
  • In the parsing experiment, frequencies were acquired from LOB plus Susanne, excluding the 250 test sentences (380K words parsed) and using add-one smoothing. The GEIG crossing-bracket score improved by 7% (p = 0.11), and the authors name filtering as the weakest component.

Interesting ideas

  • I1. Process the corpus once. Tag, lemmatize and parse the whole corpus a single time, extract patterns for every predicate in each sentence, then regroup them by predicate. This is much cheaper than re-processing the corpus per word and suits any per-lemma lexical statistics task.
  • I2. Use the classifier as a noise filter. Requiring every extracted pattern to map onto a known class silently discards many spurious shallow parses. Validating noisy parser output against a closed target inventory is a cheap form of quality control.
  • I3. Grow the inventory from rejects. New classes were found by examining the patterns the classifier could not place. Reviewing what a system rejects is a practical way to expand a taxonomy.
  • I4. Evaluate extrinsically. The acquired lexicon was judged both against reference entries and by whether it improved a downstream parser. This shows whether a resource actually helps a task, not just whether it matches a gold list.
  • I5. Keep head lemmas for later. Storing argument heads alongside each pattern allows selectional preferences or alternations to be learned later without re-parsing. This kind of forward-compatible data design makes resources easier to extend.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Natural language processing

Cited by

118 papers of our library cite it (of the 140884 whose bibliographies we read); 3 of them have a page here, 115 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.3115/974557.974609

@inproceedings{briscoe1997automatic,
  title = {{Automatic Extraction of Subcategorization from Corpora}},
  author = {Briscoe, Ted and Carroll, John},
  booktitle = {Fifth Conference on Applied Natural Language Processing},
  year = {1997},
  doi = {10.3115/974557.974609},
  url = {https://doi.org/10.3115/974557.974609}
}

References

33 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Boguraev, B. & Briscoe, E. 1987. Large lexicons for natural language processing: utilising the grammar coding system of the Longman Dictionary of Contemporary English. Computational Linguistics 13.4: 219-240. find
  2. Boguraev, B. & Briscoe, E. 1989. Introduction. In Boguraev, B. & Briscoe, E. eds. Computational Lexicography for Natural Language Processing. Longman, London: 1-40. find
  3. Boguraev, B., Briscoe, E., Carroll, J., Carter, D. & Grover, C. 1987. The derivation of a grammatically-indexed lexicon from the Longman Dictionary of Contemporary English. In Proceedings of the 25th Annual Meeting of the Association for Computational Linguistics, Stanford, CA. 193-200. find our card
  4. Brent, M. 1991. Automatic acquisition ofsubcatego-rization frames from untagged text. In Proceedings of the 29th Annual Meeting of the Association for Computational Linguistics, Berkeley, CA. 209-214. find
  5. Brent, M. 1993. From grammar to lexicon: unsupervised learning of lexical syntax. Computational Linguistics 19.3: 243-262. find
  6. Briscoe, E. & Carroll, J. 1993. Generalised probabilistic LR parsing for unification-based grammars. Computational Linguistics 19.1: 25-60. find
  7. Briscoe, E. & Carroll, J. 1994. Parsing (with) punctuation. Rank Xerox Research Centre, Grenoble, MLTT-TR-007. find
  8. Briscoe, E. & Carroll, J. 1995. Developing and evaluating a probabilistic LR parser of part-of-speech and punctuation labels. In Proceedings of the ~th A CL/SIGPARSE International Workshop on Parsing Technologies, Prague, Czech Republic. 48-58. find
  9. Carroll, J. 1993. Practical unification-based parsing of natural language. Cambridge University Computer Laboratory, TR-224. find
  10. Carroll, J. 1994. Relating complexity to practical performance in parsing with wide-coverage unification grammars. In Proceedings of the 32nd Annual Meeting of the Association for Computational Linguistics, NMSU, Las Cruces, NM. 287-294. find
  11. Carroll, J. & Briscoe, E. 1996. Apportioning development effort in a probabilistic LR parsing system through evaluation. In Proceedings of the ACL SIGDAT Conference on Empirical Methods in Natural Language Processing, University of Pensylvania, Philadelphia, PA. 92-100. find
  12. Carroll, J. & Grover, C. 1989. The derivation of a large computational lexicon for English from LDOCE. In Boguraev, B. and Briscoe, E. eds. Computational Lexicography for Natural Language Processing. Longman, London: 117-134. find
  13. Cunningham, H., Gaizauskas, R. & Wilks, Y. 1995. A general architecture for text engineering (GATE) - a new approach to language R~4D. Research memo CS-95-21, Department of Computer Science, University of Sheffield, UK. de Marcken, C. 1990. Parsing the LOB corpus. In Proceedings of the 28th Annual M… find
  14. Elworthy, D. 1994. Does Baum-Welch re-estimation help taggers?. In Proceedings of the ~th Conf. find
  15. Applied NLP, Stuttgart, Germany. Garside, R., Leech, G. & Sampson, G. 1987. The computational analysis of English: A corpus-based approach. find
  16. Longman, London. Grishman, R., Macleod, C. & Meyers, A. 1994. Comlex syntax: building a computational lexicon. In Proceedings of the International Conference on Computational Linguistics, COLING-94, Kyoto, Japan. 268-272. find
  17. Grishman, R., Macleod, C. & Sterling, J. 1992. find
  18. Evaluating parsing strategies using standardized parse files. In Proceedings of the 3rd A CL Conference on Applied Natural Language Processing, Trento, Italy. 156-161. find
  19. Grishman, R. & Sterling, J. 1992. Acquisition of selectional patterns. In Proceedings of the International Conference on Computational Linguistics, COLING-92, Nantes, France. 658-664. find
  20. Jackendoff, R. 1977. X-bar syntax. MIT Press; Cambridge, MA.. find
  21. Jensen, K. 1991. A broad-coverage natural language analysis system. In M. Tomita eds. Current Issues in Parsing Technology. Kluwer, Dordrecht. find
  22. Levin, B. 1993. Towards a lexical organization of English verbs. find
  23. Chicago University Press, Chicago. Manning, C. 1993. Automatic acquisition of a large subcategorisation dictionary from corpora. In Proceedings of the 31st Annual Meeting of the Association for Computational Linguistics, Columbus, Ohio. 235-242. find
  24. Meyers, A., Macleod, C. & Grishman, R. 1994. Standardization of the complement adjunct distinction. find
  25. New York University, Ms. Nunberg, G. 1990. The linguistics of punctuation. CSLI Lecture Notes 18, Stanford, CA. find
  26. Poznanski, V. & Sanfilippo, A. 1993. Detecting dependencies between semantic verb subclasses and subcategorization frames in text corpora. In Proceedings of the SIGLEX ACL Workshop on the Acquisition of Lexical Knowledge from Text, Boguraev, B. & Pustejovsky, J. eds. find
  27. Resnik, P. 1993. Selection and information: a class-based approach to lexical relationships. University of Pennsylvania, CIS Dept, PhD thesis. find
  28. Ribas, P. 1994. An experiment on learning appropriate selection restrictions from a parsed corpus. In Proceedings of the International Conference on Computational Linguistics, COLING-94, Kyoto, Japan. find
  29. Sampson, G. 1995. English for the computer. Oxford, UK: Oxford University Press. find
  30. Schabes, Y. 1992. Stochastic lexicalized tree adjoining grammars. In Proceedings of the International Conference on Computational Linguistics, COLING-92, Nantes, France. 426-432. find
  31. Taylor, L. & Knowles, G. 1988. Manual of information to accompany the SEC corpus: the machine-readable corpus of spoken English. University of Lancaster, UK, Ms. find
  32. Ushioda, A., Evans, D., Gibson, T. & Waibel, A. 1993. The automatic acquisition of frequencies of verb subcategorization frames from tagged corpora. find
  33. In Boguraev, B. & Pustejovsky, J. eds. SIGLEX A CL Workshop on the Acquisition of Lexical Knowledge from Text. Columbus, Ohio: 95-106. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/briscoe-1997-automatic-extraction-subcategorization-corpora/