Grammatical Framework (GF) for MT in Sublanguage Domains

What it is about

A description of the GF open-source package for building interlingua MT applications in narrow domains. A grammar has two parts: an abstract syntax, which is language-independent and lists the categories and functions of the domain, and a concrete syntax for each language, which maps the functions to strings. Concrete syntaxes are built on a resource grammar library of ten languages, so that an application author needs only the domain semantics and the lexicon. The paper shows a small health phrase-book that translates simple sentences among four languages through a shared interlingua, and it describes an editor that parses an example sentence and pre-fills the rule from the library. It treats the expressive power of the formalism: rules can permute, suppress and duplicate arguments, which lets the grammar handle word order and articles that differ between languages. The paper also shows a case where the interlingua must ask the author for a feature, the gender of the pronoun I, because Russian agreement needs it.

Ideas

I1. Mark a feature as unknown instead of guessing it. When the reader cannot decide a feature that some target language needs (gender, number, person), the meta token keeps the value as unknown, and the encoder uses a stated default per author and the residual records the guess. This follows open problems 5 and 4 and makes the guess visible in the residual.
I2. Propose a delta from one example pair and count it as learned only after it passes the existing gate, which is at least three more tokens fixed. The editor's prefill from one example is the model; the gate is the difference.
I3. Make the article and determiner handling explicit as suppression on chunks. English and Hebrew differ in the article, so an encoder rule that drops or adds a determiner should be one named operation, counted per author, not an effect of the chunk string.

For Mova

The code is tools/translate (meta.rs for the feature state of meta tokens, profile.rs for the encoder's chunk operations) and the Hebrew reader in hebrew.rs. The formalism is not for Mova, which keeps its grammar in trees and counts rather than in a hand-written abstract syntax. The unknown-feature state (I1) is the idea to test first, on the gender and number of the Hebrew nouns.

Topics: Machine translation

Cite

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

@inproceedings{khegai2006grammatical,
  title = {{Grammatical Framework (GF) for MT in Sublanguage Domains}},
  author = {Khegai, Janna},
  booktitle = {Proceedings of the 11th Annual Conference of the European Association for Machine Translation},
  year = {2006},
  url = {https://aclanthology.org/2006.eamt-1.13/}
}

References

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

Bringert, B., Cooper, R., Ljunglöf, P., & Ranta, A. (2005). Multimodal Dialogue System Grammars. In DIALOR’05, Ninth Workshop on the Semantics and Pragmatics of Dialogue, Nancy, France. find
Burke, D., & Johannisson, K. (2005). Translating Formal Software Specifications to Natural Language / A Grammar-Based Approach. In J. B. P. Blace, E. Stabler & R. Moot (Eds.), Logical Aspects of Computational Linguistics (LACL 2005) (Vol. 3402, pp. 51–66). Springer. Butt, M., King, T. H., no, M.-E. find
N., & Segond, F. (Eds.). (1999). A grammar writer’s cookbook. Stanford: CSLI Publications. find
Caprotti, O. (2006). WebALT! Deliver Mathematics Everywhere. In SITE 2006, Orlando, USA. (webalt.math.helsinki.fi/content/ e16/e301/e512/PosterDemoWebALT.pdf) find
Cooper, R., & Ranta, A. (2004). Dialogue systems as proof editors. The Jornal of Logic, Language and Information. ↗
Copestake, A., & Flickinger, D. (2000). An open-source grammar development environment and broad-coverage english grammar using hpsg. In Second conference on Language Resources and Evaluation (LREC-2000), Athens, Greece. find
Crouch, D., Dalrymple, M., Kaplan, R., King, T., Maxwell, J., & Newman, P. (2005). XLE documentation. (URL: www2.parc.com/istl/groups/nltt/xle) find
Dymetman, M., Lux, V., & Ranta, A. (2000). XML and multilingual document authoring: Convergent trends. In COL-ING, Saarbrücken, Germany (pp. 243– 249). find
Hähnle, R., Johannisson, K., & Ranta, A. (2002). An authoring tool for informal and formal requirements specifications. In R.-D. Kutsche & H. Weber (Eds.), Fundamental Approaches to Software Engineering (Vol. 2306, pp. 233– 248). Springer. find
Jurafsky, D., & Martin, J. (2000). Speech and language processing. Prentice Hall. find
Khegai, J. (2005). GF IDE for GF 2.1. www.cs.chalmers.se/~aarne/GF2.0/GF-Doc/GF_IDE_manual/index.htm. find
Khegai, J., Nordström, B., & Ranta, A. (2003). Multilingual syntax editing in GF. In A. Gelbukh (Ed.), CICLing-2003, Mexico City, Mexico (pp. 453–464). Springer. find
Lindström, J. (1995). Summary on reduplication. LINGUIST List: Vol-6-52. find
Ljunglöf, P. (2004). Expressivity and Complexity of the Grammatical Framework. (URL: www.cs.chalmers.se/~peb/pubs/p04-PhD-thesis.pdf) find
Magnusson, L., & Nordström, B. (1994). The ALF proof editor and its proof engine. In Types for Proofs and Programs (pp. 213–237). Springer. find
Pollard, C., & Sag, I. (1994). Head-Driven Phrase Structure Grammar. University of Chicago Press. find
Power, R., Scott, D., & Evans, R. (1998). Generation as a solution to its own problem. In Inlg’98. Niagara-on-the-Lake, Canada. find
Ranta, A. (2004). Grammatical Framework: A Type-theoretical Grammar Formalism. The Journal of Functional Programming, 14 (2), 145–189. find
Ranta, A. (2006). GF Homepage. (www.cs.chalmers.se/~aarne/GF/) find
Ranta, A. (to appear, 2005). Modular Grammar Engineering in GF. Research in Language and Computation. (URL: www.cs.chalmers.se/~aarne/articles/ar-multieng.pdf) find
Rayner, M., Carter, D., Bouillon, P., Digalakis, V., & Wirén, M. (2000). The spoken language translator. Cambridge University Press. ↗

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/khegai-2006-grammatical-framework-gf-mt-sublanguage/