Grammatical Framework: an Interlingual Grammar Formalism

What it is about

A two-page tutorial outline for Grammatical Framework (GF), a grammar formalism started in 1998 for interlingual translation. A GF grammar has an abstract part (the interlingua, built on a constructive type theory) and one concrete part per language. The same rules serve to analyse and to generate, so a mapping can be run in both directions. The concrete rules are equivalent to parallel multiple context-free grammars. Community grammars cover more than 40 languages. The paper also says that GF has been combined with neural dependency parsing to scale up coverage while keeping the analysis explainable. There are no experiments: the text lists a 45-minute hands-on session, a theory part and a survey of the ecosystem.

Ideas

• I1. Reversible tables. Today the reader's lookups and the standard encoder's tables in tools/translate are built as separate structures. Try one table of concept, tag and chunk that is used for reading and for writing. Count the pairs where the two directions disagree on the Bible pairs. A disagreement is a candidate for a merge or split in profile.rs.
• I2. Abstract versus concrete split. Keep the meta-world as an order-free abstract syntax, and let each target language own its linearisation. Measure the residual that is due to order alone, per target, to see how much en/src/lin.rs has to explain.
• I3. Explicit coverage. A GF grammar either analyses a sentence or rejects it. Report the Hebrew reader's constructions that have no rule, as a coverage number, instead of a silent best guess.

For Mova

The paper has no measurement, so the ideas are design tests, not findings. They bear on open problem 3 (the encoder should write trees, not word strings) in tools/translate/src/profile.rs, and on the reversibility of the lexicon in meta.rs. The GF rules are hand-written, while Mova's tables are counted from data. That trade-off should be measured, not assumed.

Topics: Machine translation

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.18653/v1/w19-3101

@inproceedings{ranta2019grammatical,
  title = {{Grammatical Framework: an Interlingual Grammar Formalism}},
  author = {Ranta, Aarne},
  booktitle = {Proceedings of the 14th International Conference on Finite-State Methods and Natural Language Processing},
  year = {2019},
  doi = {10.18653/v1/w19-3101},
  url = {https://doi.org/10.18653/v1/w19-3101}
}

References

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

1. Krasimir Angelov, The Mechanics of the Grammatical Framework, PhD Thesis, Chalmers University of Technology, 2011. Standard reference on the internals of GF, from both implementation and formal language theory point of view. find
2. Peter Ljunglöf, Expressivity and Complexity of the Grammatical Framework, PhD Thesis, University of Gothenburg, 2004. Detailed study of the language-theoretic properties of GF, placing it among mildly context-sensitive formalisms. find
3. Aarne Ranta, Grammatical Framework: Programming with Multilingual Grammars, CSLI, Stanford, 2011. Standard reference on the GF programming language, with a tutorial and reference manual. find
4. Aarne Ranta, “What are Grammars Good for? Reflections on Twenty Years of Grammatical Framework”, in Cleo Condoravdi and Tracy Holloway King (eds), Tokens of Meaning. Papers in Honor of Lauri Karttunen, CSLI, Stanford, 2019, pp. 545–568. Follow-up of recent research themes, with some parallels to Xerox Finite State Tool, which was an important inspiration for GF. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/ranta-2019-grammatical-framework-interlingual-grammar-formalism/