Using Danish as a CG Interlingua: A Wide-Coverage Norwegian-English Machine Translation System

What it is about

The paper builds a rule-based Norwegian-English system and uses Danish as a pivot, the "interlingua" of the title. Norwegian and Danish are close, so a Norwegian-Danish lexicon is cheap to build, and a Danish-English system already works well. Parsing and polysemy resolution use Constraint Grammar function tags and dependency structures. The main work is the Norwegian-Danish dictionary, built semi-automatically from about 226,000 lemma candidates, mostly by spelling rules and comparison with existing word lists. To check the method, the authors inspect all 1,544 translation candidates for words of frequency 9. About 12 percent of the unrevised candidates are wrong, and most of those errors (10.4 percent) are Danish words that are etymologically close to the Norwegian one. The authors also note that a one-to-one lexicon is enough for close languages, and that many-to-many lexicons are needed for less related pairs.

Ideas

• I1. Use a close author as a pivot for another: read the meta-world of a close translation (for example WEB or ASV) as the source for a further author, and compare the cost of that delta with the cost of a delta from the standard. The author delta code already has the structure for this.
• I2. Audit the lexicon the same way: sample a fixed number of frequent merges, inspect them, and report the share that are noise. The paper's error rate of about 12 percent is a useful baseline to beat. Run the audit after each merge round.
• I3. Disambiguate a concept by its head relation, as the paper does with function tags. The reader's relation then becomes an input to the choice of concept.
• I4. Keep merges one to one where possible, and many-to-many only with a context key. Measure how many of the merges kept by the current rules are many-to-many.

For Mova

The places are tools/translate/meta.rs (the merges and the audit), tools/translate/hebrew.rs (the alignment that produces the pairs) and profile.rs (context keys). I2 is the lexicon test of open problem 2. I3 is open problem 1 (the reader's relations). The pivot idea in I1 is a close relative of the delta design, and the paper gives no measures beyond the audit.

Topics: Machine translation

Cite

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

@inproceedings{bick2007using,
  title = {{Using Danish as a CG Interlingua: A Wide-Coverage Norwegian-English Machine Translation System}},
  author = {Bick, Eckhard and Nygaard, Lars},
  booktitle = {Proceedings of the 16th Nordic Conference of Computational Linguistics (NODALIDA 2007)},
  year = {2007},
  url = {https://aclanthology.org/W07-2405/}
}

References

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

1. Bick, Eckhard. 2001. «En Constraint Grammar Parser for Dansk», in Peter Widell & Mette Kunøe (eds.), 8. Møde om Udforskningen af Dansk Sprog, 12.-13. oktober 2000, pp. 40-50, Århus University Bick, Eckhard. 2003, «A CG & PSG Hybrid Approach to Automatic Corpus Annotation», In: Kiril Simow & Petya Osenova (eds.), Proceedings of SProLaC2003 (at Corpus Linguistics 2003, Lancaster), pp. 1-12 find
2. Bick, Eckhard. 2005 «Turning Constraint Grammar Data into Running Dependency Treebanks», In: Civit, Montserrat & Kübler, Sandra & Martí, Ma. Antònia (red.), Proceedings of TLT 2005 (4th Workshop on Treebanks and Linguistic Theory, Barcelona, December 9th - 10th, 2005), pp.19-27 find
3. Bick, Eckhard. 2006. «A Constraint Grammar Based Spellchecker for Danish with a Special Focus on Dyslexics». In: Suominen, Mickael et al. (ed.) A Man of Measure: Festschrift in Honour of Fred Karlsson on his 60th Birthday. Special Supplement to SKY Jounal of Linguistics, Vol. 19 (ISSN 1796-279X), pp. 387-396. Turku: The Linguistic Association of Finland Bick, Eckhard. 2007. «Fra syntaks til semantik: Polysemiresolution igennem dependensstrukturer i dansk-engelsk maskinoversættelse.» (forthcoming) find
4. Corbí-Bello, Antonio M. et al. 2005. An open-source shallow-transfer machine translation engine for the Romance Languages of Spain. In Proceedings of the European Association for Machine Translation, 10th Annual Conference, Budapest 2005, p. 79-86. find
5. Hagen, Kristin, Johannessen, Janne Bondi, Nøklestad, Anders. 2000. "A Constraint-Based Tagger for Norwegian". In: Lindberg, C.-E. and Lund, S.N. (red.): 17th Scandinavian Conference of Linguistic, Odense. Odense Working Papers in Language and Communication, No. 19, vol I. find
6. Karlsson, Fred. 1990. Constraint Grammar as a Framework for Parsing Running Text. In: Karlgren, Hans (ed.), COLING-90 Helsinki: Proceedings of the 13th International. Conference on Computational Linguistics, Vol. 3, pp. 168-173 find
7. Lønning, Jan Tore, Stephan Oepen, Dorothee Beermann, Lars Hellan, John Carroll, Helge Dyvik, Dan Flickinger, Janne Bondi Johannessen, Paul Meurer, Torbjørn Nordgård, Victoria Rosén, and Erik Velldal. 2004. LOGON. A Norwegian MT effort. In Proceedings of the Workshop in Recent Advances in Scandinavian Machine Translation, Uppsala, Sweden, Paul, Michael. 2001. Knowledge Recycling for Related Languages. Proceedings of MT Summit VIII. Santiago de Compostela, Spain. pp. 265-269. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/bick-2007-danish-as-cg-interlingua-wide/