Elamur Research⌕
☰
home / papers / 1993 / A Principle-based Parser for Foreign Language Training in German and Arabic

A Principle-based Parser for Foreign Language Training in German and Arabic

Paper · 1993 · Joe Garman, Jeffery Martin, Paola Merlo et al.

Summary

The authors build a parser for a computer-based foreign language tutor that has to cope with learner input, which is often misspelled or ungrammatical. The tutor also runs on PCs with little spare memory. The parser covers German and Arabic, with Spanish planned. It follows Government-Binding theory: a small, maximally binary X-bar context-free backbone is driven by an augmented shift-reduce algorithm. Each rule is tied to a conjunction of feature constraints, and a separate move-x step builds gaps at parse time for questions, relatives and topicalisation. A morphological analyser supplies every possible reading of each word. An interactive preprocessor expands contractions and asks the student to fix misspellings. Grammar violations such as agreement, case, selection, auxiliary and word order are handled by default mechanisms, so the parser still produces a tree together with a list of diagnosed errors for the tutor. On 205 German test sentences written by educators, about 96% received a parse, including ill-formed ones. The work shows that factoring grammar into principles yields a compact, portable and error-tolerant system.

Key points

  • The output is a parse tree, which goes to a semantic analyser, plus a possibly empty list of errors, which goes to the tutor. Students never see the trees.
  • German test batch of 205 sentences: 135 of 141 correct sentences parsed (96.4%), 61 of 64 incorrect sentences parsed (95.3%), and 196 overall (95.6%).
  • Lexicons hold 5000 German roots and 500 Arabic roots. The parsers are 1632 (German) and 1736 (Arabic) lines of Prolog.
  • An interpreted version on a SUN workstation ran at about 14 words/s for German and 27 words/s for Arabic, roughly as fast as a human tutor gives feedback.
  • Arabic input is currently in phonetic transcription. Arabic error types centre on nominal agreement, plural formation and noun construct, plus subject-verb agreement.

Interesting ideas

  • I1. Keep phrase-structure rules separate from the feature constraints that license them, so a failed constraint can be logged as a named diagnosis instead of killing the parse
  • I2. Treat constraint modules as switchable, giving a strict mode that reports every violation and a lenient mode that ignores feature mismatches
  • I3. Have the morphology return the full set of analyses per word and let later syntactic constraints choose, avoiding early commitment and backtracking failures
  • I4. Handle the two error classes differently: repair spelling up front, but absorb grammatical errors during parsing via defaults
  • I5. Evaluate on an externally designed test suite that mixes correct and deliberately erroneous sentences, and report each group separately

For Mova

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

  • I1 (tree): the parser could attach agreement or selection violations as explicit, explainable annotations on an otherwise valid UD tree, which is useful for noisy or learner text.
  • I2 (answer): readers answering school tests could run leniently on student-written input but strictly when checking well-formedness, with the same grammar underneath.
  • I3 (tokens): the en tagger could pass ambiguous lemma/feature sets forward so the parser can resolve them with an explicit reason.
  • I4 (tokens): a pre-parse normalisation layer for typos and contractions, kept apart from syntactic tolerance, would make each repair traceable to one stage.
  • I5 (learn): ill-formed sentences make a natural negative-control set, showing whether robustness on bad input comes at the cost of accuracy on clean input.

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, 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

🧭 Topics 1Research subjects and languages the work is about.Semitic languages · 2,908 works

Cite

Cite as Joe Garman, Jeffery Martin, Paola Merlo and Amy Weinberg. 1993. A Principle-based Parser for Foreign Language Training in German and Arabic. Proceedings of the Third International Workshop on Parsing Technologies.

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

@inproceedings{garman1993principle,
  title = {{A Principle-based Parser for Foreign Language Training in German and Arabic}},
  author = {Joe Garman, Jeffery Martin, Paola Merlo and Amy Weinberg},
  booktitle = {Proceedings of the Third International Workshop on Parsing Technologies},
  year = {1993},
  url = {https://aclanthology.org/1993.iwpt-1.8/}
}

References

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

  1. Aho, A.V. - J.D. Ullman (1972) The Theory of Parsing, Translation and Compiling. Engle wood Cliffs, NJ: Prentice-Hall. find
  2. Azadegan, S. - J. Martin - P. Merlo - A. Weinberg (forthcoming) A Government Binding Parser for Foreign Language Train ing. UMIACS-TR, College Park, MD: UMCP. Berwick, R. ( 1991 ) "Principle-Based Parsing" . In: Sells P. & S.M. Shieber & T. Wasow (Eds.) : Foundational Issues in Natural Lan guage Processing . 1 1 5-226. Cambridge, MA: MIT Press. find
  3. Berwick, R. - A. Weinberg (1984) The Gram matical Basis of Linguistic Performance. Cambridge, MA: MIT Press. find
  4. Chomsky, N. ( 1965) Aspects of the Theory of Syntax . Cambridge, MA: MIT Press. find
  5. Chomsky, N. (1970) "Remarks on Nominal ization" . In: Jacobs R. & P. Rosenbaum (Eds.): Readings in Transformational Gram mar. 184-221 . Waltham, MA: Ginn & Co. find
  6. Chomsky, N. ( 1973) "Conditions on Transforma tions" . In: Anderson S. & P. Kiparsky (Eds.): A Festschrift for Morris Halle . 232-286. New York, NY: Holt, Reinhart and Winston. find
  7. Chomsky, N. ( 1977) "On Wh-movement" . In: Culicover P. & T. Wasow & A. Akmajian (Eds.): Formal Syntax . 71-132. New York, NY: Academic Press. find
  8. Chomsky, N. (1981 ) Lectures on Government and Binding. Dordrecht: Foris. find
  9. Crocker, M.W. (1993) "Properties of the Principle-Based Sentence Processor" . In: Proceedings of the 15th Annual Cognitive Sci ence Society . Boulder, CO. find
  10. Dorr, B.J. (1987) UNITRAN: a Principle-based Approach to Machine Translation: Ms Thesis, MIT. Cambridge, MA. find
  11. Dorr, B.J. (1993) Machine Translation: A View from the Lexicon . Cambridge, MA: MIT Press.
  12. Fong, S. ( 1991) Computational Properties of Principle-based Grammatical Theories . Ph.D. Dissertation, MIT. Cambridge, MA. find
  13. Gazdar G. - E.Klein - G.Pullum - I.Sag (1985) Generalized Phrase Structure Gram mar. Oxford: Blackwell. find
  14. Jackendoff, R. (1977) X Syntax: A Study of Phrase Structure . Cambridge, MA: MIT Press. find
  15. Kayne, R. (1984) Connectedness and Binary Branching . Dordrecht: Foris. find
  16. Kornai, A. - G. Pullum (1990) "The X-bar The ory of Phrase Structure" . In: Language 66, 24 - 50. find
  17. Merlo, P. ( 1992) On Modularity and Compilation in a Government and Binding Parser. Ph.D. . Dissertation, University of Maryland at Col lege Park. College Park, MD . find
  18. Shieber, S. (1986) An Introduction to Unification-Based Approaches to Grammar. Chicago, IL: University of Chicago Press. find
  19. Thiersch, C. (1978) Topics in German Syntax, Ph.D . Dissertation, MIT. Cambridge, MA. find
  20. Tomita, M. (1985) Efficient Parsing for Natural Language. Ringham, MA: Kluwer. ↗
  21. Wanner, E. - M. Maratsos (1978) "An ATN Approach to Comprehension" In: Halle M. & J. Bresnan & G.A. Miller (Eds.): Linguistic Theory and Psychological Reality. 119-161. Cambridge, MA: MIT Press. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/garman-1993-principle-based-parser-foreign-language/