Three studies of grammar-based surface-syntactic parsing of unrestricted English text. A summary and orientation

What it is about

The dissertation argues that a hand-written, rule-based grammar can tag unrestricted English as accurately as the statistical taggers of the time, if it is allowed to leave the hardest ambiguities pending. The English Constraint Grammar has about 1,100 disambiguation constraints, which express 23 general syntactic statements about the linear order of morphological tags. On a 2,167-word test it reaches 99.77% recall at about 1.04 readings per word, against 96.95% for CLAWS1 and 96.21% for the Church tagger, which both give one reading per word. A further 200 heuristic constraints resolve about half of what is left, with 96% of those predictions correct. Three experiments follow. Lexical probabilities learned from the same text resolve 36% of the noun-verb ambiguities left pending, at 95.5% accuracy. Noun groups extracted from 215,000 words resolve about 27% of the surviving ambiguities. A last chapter proposes one finite-state grammar for morphology, clause boundaries and syntax, and resolves 40 of 43 hard cases on a 1,400-word text.

Ideas

• I1. Table-driven constraints in en/src/tag.rs: write each of the 23 generalisations as a negative constraint over the tag sequence (for example, a determiner followed by a finite verb reading is removed). Store them in a BTreeMap keyed by the constraint name, so their application order is fixed, and report which constraints fire on which lesson.
• I2. Report the paper's two numbers together: recall and readings per word. Add them to en parse-eval and tools/parsecmp, so that a system that keeps more ambiguity is not credited for it.
• I3. Text-based lexical probabilities: take the words that the first pass fully disambiguates in a school text, and use their tag counts to decide a pending noun-verb case, but only if one tag is at least twice as frequent as the other. Measure accuracy on the pending cases, with a negative control that uses a 1:1 threshold.
• I4. Noun-group propagation in en/src/ptag.rs: once a noun group is resolved in one place in a text, mark all its other occurrences. Check the reliability by hand on a random sample, as the paper does with 2,000 sequences.
• I5. Morphological heuristics for unknown words in en/src/morph.rs: ordered rules on the form of the word (ending, prefix) that propose a reading, with a flag for the rule that made it. Measure the accuracy on words missing from the lexicon.

For Mova

Tagging and ambiguity: 10. Word classes and 11. Nouns (noun-verb ambiguity), 13. Verbs. Unknown words and derivation: 18. Compounding, conversion and the minor processes, 16. Derivation I: prefixes. Code: en/src/tag.rs and en/src/ptag.rs (constraints, text-based probabilities), en/src/morph.rs (heuristics for unknown words), en parse-eval and tools/parsecmp (recall at a fixed ambiguity level). Measure: recall and readings per word on a held-out school text, run twice for determinism, and the per-constraint hit counts.

Cite

The paper itself, not this card. BibTeX from the library's record of the work. arXiv cmp-lg/9406039

@misc{voutilainen1994three,
  title = {{Three studies of grammar-based surface-syntactic parsing of unrestricted English text. A summary and orientation}},
  author = {Voutilainen, Atro},
  year = {1994},
  eprint = {cmp-lg/9406039},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/cmp-lg/9406039}
}

References

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

1. Black, E. 1993. Statistically-Based Computer Analysis of English. In Black & al. (eds.), 1-16. find
2. Black, E., Garside, R. and Leech, G. (eds.) 1993. Statistically-Driven Computer Grammars of English: The IBM/Lancaster Approach . Amsterdam: Rodopi. find
3. Church, K. 1988. A Stochastic Parts Program and Noun Phrase Parser for Unrestricted Text. Proceedings of the Second Conference on Applied Natural Language Processing, ACL . 136-143. find
4. Church, K. 1992. Current Practice in Part of Speech Tagging and Suggestions for the Future, in Simmons (ed.), Sbornik praci: In Honor of Henry Kucera . Michigan Slavic Studies. find
5. Eyes, E. and Leech, G. 1993. Syntactic Annotation: Linguistic Aspects of Grammatical Tagging and Skeleton Parsing. In Black & al. (eds.), 36-61. find
6. Garside, R. and McEnery, A. 1993. Treebanking: The Compilation of a Corpus of Skeleton-Parsed Sentences. In Black & al. (eds.), 17-35. find
9. Heikkila, J. 1994. A TWOL-Based Lexicon and Feature System for English. In Karlsson & al. find
10. Heikkila, J. 1994. ENGTWOL English lexicon: solutions and problems. In Karlsson & al. find
11. Hindle, D. 1983. Deterministic Parsing of Syntactic Non- uencies. Proceedings of the 21st Annual Meeting of the ACL . 123-128. find
12. Karlsson, F. 1985. Parsing Finnish in terms of Process Grammar. In F. Karlsson (ed.) Computational Morphosyntax. Report on Research 1981-1984 . Publication No. 13, Department of General Linguistics, University of Helsinki. find
13. Karlsson, F. 1990. Constraint Grammar as a Framework for Parsing Running Text. In H. Karlgren (ed.), Papers presented to the 13th International Conference on Computational Linguistics, Vol. 3 . Helsinki. 168-173. find
14. Karlsson, F. 1994. Designing a parser for unrestricted text. In Karlsson & al. [Karlsson, Voutilainen, Heikkila and Anttila (eds.), 1994] Karlsson, F., Voutilainen, A., Heikkila, J. and Anttila, A. (eds.) 1994. Constraint Grammar: a Language-Independent System for Parsing Unrestricted Text . Berlin and New York: Mouton de Gruyter. find
15. Koskenniemi, K. 1990. Finite-state parsing and disambiguation. In Karlgren, H. (ed.) COLING{90. Papers presented to the 13th International Conference on Computational Linguistics, Vol. 2 . Helsinki, Finland. 229{ 232. [Koskenniemi, Tapanainen and Voutilainen, 1992] Koskenniemi, K., Tapanainen, P. and Voutilainen, A. 1992. Compiling and using nite-state syntactic rules. In Proceedings of the fteenth International Conference on Computational Linguistics. COLING-92, Vol. I . Nantes, France. 156{162. find
16. Marcken, C. de 1990. Parsing the LOB Corpus. Proceedings of the 28th Annual Meeting of the ACL . 243-251. find
17. Marshall, I. 1983. Choice of grammatical word-class without global syntactic analysis: tagging words in the LOB Corpus. Computers in the Humanities 17. 139{150. [Quirk, Greenbaum, Leech and Svartvik, 1985] Quirk, R., Greenbaum, S., Leech, G. and Svartvik, J. 1985. A Comprehensive Grammar of the English Language . London & New York: Longman. find
18. Sampson, G. 1987. Probabilistic Models of Analysis. In Garside, Leech & Sampson (eds.) 1987. 16-29. find
19. Sampson, G. 1987. The grammatical database and parsing scheme. In Garside, Leech and Sampson (eds.) 1987. 82-96. find
20. Sampson, G. (forthcoming). The need for grammatical stocktaking. To appear in N. Ostler (ed), Proceedings of the 1992 Pisa Symposium on European Textual Corpora . find
21. Tapanainen, P. 1991. A  arellisina automaatteina esitettyjen kielioppisaantojen soveltaminen luonnollisen kielen jasentajassa [Natural language parsing with nite-state syntactic rules]. Master's thesis. Department of computer science, University of Helsinki. find
22. Voutilainen, A. 1994. A syntax-based part-of-speech disambiguator. A manuscript submitted to ACL'94. find
23. Voutilainen, A. and Heikkila, J. 1994. Compiling and testing the lexicon. In Karlsson & al. find
24. Voutilainen, A. and Heikkila, J. 1994. ENGCG: a fast and accurate Constraint Grammar parser of English. In Tottie and Fries (eds.), Proceedings of ICAME'93 . Amsterdam and Atlanta: Rodopi. find
25. Voutilainen, A. and Tapanainen, P. 1993. Ambiguity resolution in a reductionistic parser. Proceedings of EACL'93 . Utrecht, Holland. 394{403. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/voutilainen-1994-three-studies-grammar-based-surface/