Automatic induction of shallow-transfer rules for open-source machine translation

What it is about

The method learns the structural transfer rules of a shallow-transfer MT system from a small parallel corpus, with no supervision. It takes alignment templates from statistical MT, in which phrase pairs are generalised over word classes, and uses as word classes the part-of-speech and inflection tags that the rest of the system already produces. Words of lexicalised categories, such as prepositions and auxiliary verbs, keep their lemma as their class. The generalised templates are then extended with restrictions taken from the bilingual dictionary: a template may apply only if the target word has the features it needs, so that a rule that would change the gender of a noun is blocked for nouns that cannot change it. Word alignments come from GIZA++ run in both directions, symmetrised. The experiments on Spanish-Catalan show a clear gain over word-for-word translation and a quality close to hand-written rules. The rules are written in the engine's XML format, so an expert can edit them.

Ideas

• I1. Use word classes built from the tags the parser already gives: coarse tag plus UD FEATS, with function words such as prepositions, determiners and auxiliaries kept as single-word classes. This is the lexicalised-category idea and it answers open problems 4 and 5, since function words and number or person become part of the meta-token class.
• I2. Guard each author delta with the target features it needs. A delta that rewrites a chunk should apply only when the standard encoder's chunk agrees in number and person with the delta's target, the same way the restrictions block gender changes. Test this as a rule over the dictionary features, not over the tokens.
• I3. Symmetrise the Hebrew aligner. The paper's intersection-then-growth step, applied to alignments in both directions, is a standard way to improve IBM Model 1 output. Use it in hebrew.rs when the alignment for open problem 7 is redone, and count how many aligned pairs change.
• I4. Keep the extraction constraint as a control: a chunk pair counts only if its first and last words are aligned on both sides. Run it with and without the constraint on one author and report the effect on the residual.

For Mova

The code is tools/translate: hebrew.rs (alignment), profile.rs (the classes and the guards on author deltas) and meta.rs. The word-class idea answers open problems 4 and 5 directly. The paper's own method is statistical and rule-based at once, so it is a model for the delta gate, not a component to copy.

Topics: Machine translation

Cite

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

@inproceedings{sanchezmartinez2007automatic,
  title = {{Automatic induction of shallow-transfer rules for open-source machine translation}},
  author = {Sánchez-Martínez, Felipe and Forcada, Mikel L.},
  booktitle = {Proceedings of the 11th Conference on Theoretical and Methodological Issues in Machine Translation of Natural Languages: Papers},
  year = {2007},
  url = {https://aclanthology.org/2007.tmi-papers.22/}
}

References

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

1. C. Armentano-Oller, R.C. Carrasco, A.M. Corbı́-Bellot, M.L. Forcada, M. Ginestı́-Rosell, S. Ortiz-Rojas, J.A. Pérez-Ortiz, G. Ramı́rez-Sánchez, F. Sánchez-Martı́nez, and M.A. Scalco. 2006. Open-source Portuguese-Spanish machine translation. In Computational Processing of the Portuguese Language, Proceedings of the 7th International Workshop on Computational Processing of Written and Spoken Portuguese, PROPOR 2006, volume 3960 of Lecture Notes in Computer Science, pages 50– 59. Springer-Verlag. find
2. D. Arnold, 2003. Computers and Translation: A translator’s guide, chapter Why translation is difficult for computers, pages 119–142. Benjamins Translation Library. Edited by H. Somers. find
3. R.D. Brown. 1999. Adding linguistic knowledge to a lexical example-based translation system. In Proceedings of the Eighth International Conference on Theoretical and Methodological Issues in Machine Translation (TMI-99), pages 22–32. find
4. H.M. Caseli, M.G.V. Nunes, and M.L. Forcada. 2005. LIHLA: A lexical aligner based on language-independent heuristics. In Proceedings of the V Encontro Nacional de Inteligência Artificial (ENIA 2005), pages 641–650. find
5. I. Cicekli and H.A. Güvenir. 2001. Learning translation templates from bilingual translation examples. Applied Intelligence, 15(1):57–76. find
6. D. Cutting, J. Kupiec, J. Pedersen, and P. Sibun. 1992. A practical part-of-speech tagger. In Third Conference on Applied Natural Language Processing. Association for Computational Linguistics. Proceedings of the Conference., pages 133–140. find
7. H. Kaji, Y. Kida, and Y. Morimoto. 1992. Learning translation templates from bilingual text. In Proceedings of the 14th Conference on Computational Linguistics, pages 672–678. Association for Computational Linguistics. find
8. A. Lavie, K. Probst, E. Peterson, S. Vogel, L. Levin, A. Font-Llitjós, and J. Carbonell. 2004. A trainable transfer-based machine translation approach for languages with limited resources. In Proceedings of Workshop of the European Association for Machine Translation (EAMT-2004). find
9. V.I. Levenshtein. 1965. Binary codes capable of correcting deletions, insertions, and reversals. Doklady Akademii Nauk SSSR, 163(4):845–848. English translation in Soviet Physics Doklady, 10(8):707-710, 1966. find
10. Y. Liu and C. Zong. 2004. The technical analysis on translation templates. In Proceedings of the IEEE International Conference on Systems, Man & Cybernetics (SMC), pages 4799–4803. IEEE. find
11. A. Mikheev. 1996. Unsupervised learning of word-category guessing rules. In Proceedings of the Thirty-Fourth Annual Meeting of the Association for Computational Linguistics, pages 327–333. find
12. F.J. Och and H. Ney. 2003. A systematic comparison of various statistical alignment models. Computational Linguistics, 29(1):19–51. find
13. F.J. Och and H. Ney. 2004. The alignment template approach to statistical machine translation. Computational Linguistics, 30(4):417– 449. find
14. F.J. Och, C. Tillmann, and H. Ney. 1999. Improved alignment models for statistical machine translation. In Proceedings of the Joint SIG-DAT Conference on Empirical Methods in Natural Language Processing and Very Large Corpora, pages 20–28. find
15. K. Probst, L. Levin, E. Peterson, A. Lavie, and J. Carbonell. 2002. MT for minority languages using elicitation-based learning of syntactic transfer rules. Machine Translation, 17(4):245–270. find
16. F. Sánchez-Martı́nez and H. Ney. 2006. Using alignment templates to infer shallow-transfer machine translation rules. In Advances in Natural Language Processing, Proceedings of 5th International Conference on Natural Language Processing FinTAL, volume 4139 of Lecture Notes in Computer Science, pages 756–767. Springer-Verlag. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/martinez-2007-automatic-induction-shallow-transfer-rules/