Elamur Research⌕
☰
home / papers / 2007 / Flexible automatic look-up of English idiom entries in dictionaries

Flexible automatic look-up of English idiom entries in dictionaries

Paper · 2007 · Koichi Takeuchi, Takashi Kanehila, Kazuki Hilao et al.

Summary

The paper addresses a practical gap in translation aids. Good dictionaries already list the idioms translators need, but neither paper dictionaries, electronic dictionaries nor MT systems reliably find an entry when the idiom appears in a modified form, for example with inserted adjectives. The authors interviewed and surveyed online English-to-Japanese volunteer translators. Translators wanted several candidate entries rather than a single answer, and wanted tolerance for variants that linguists would call impossible.

The authors then built a dataset of idiom variants. Three native speakers rewrote entries from an English-Japanese idiom dictionary as creatively as possible. The resulting variants were classified, and syntagmatic augmentation (insertions) turned out to be the dominant type. The lookup system works in two steps. First, a recall-oriented surface matcher runs over POS-tagged, lemmatised text, without a parser. Second, POS-based filtering rules remove matches that do not fit known variation patterns. On real-world text, filtering lifted precision from about 0.15 to about 0.53 while recall stayed near 0.98. The authors consider this good enough for real translator use.

Key points

  • Translator requirements: eight online translators were consulted in person, and 12 more answered a questionnaire. Two requirements emerged:

    • return multiple candidates;
    • accept liberal, "unlinguistic" variants, which implies aiming for recall near 100%.

  • Variant data: informants produced 2181 variants for 1913 idioms taken from McCaleb and Iwasaki (2003). Syntagmatic augmentation was the largest class (1203 cases), followed by paradigmatic replacement (759).

  • Four-way variant taxonomy:

    1. dictionary-registered type variants;
    2. externally driven syntactic operations, such as passivisation and topicalisation;
    3. internal insertions or replacements;
    4. highly creative wordplay. The work targets class 3.

  • Results:

    Data setMethodPrecisionRecall
    (a) constructed datasurface matching only0.4180.991
    (a) constructed datamatching + filtering0.7340.968
    (b) real-world datasurface matching only0.1470.996
    (b) real-world datamatching + filtering0.5280.978

  • Error sources:

    • lemmatisation and POS-tagging mistakes (e.g. "damned" tagged as an adverb, a mis-tagged entry "from high");
    • unconstrained placeholders in entries such as "make A of B";
    • long insertions that the patterns miss (e.g. "takes her daughter-in-law apart").

Interesting ideas

  • I1. Overgenerate, then filter: match loosely to keep recall almost perfect, then prune with cheap pattern rules; this keeps the error profile tunable and inspectable.
  • I2. Elicited variants instead of corpus mining: ask native speakers (or a teacher model) to invent creative variants, so the data maps the tractable range of variation rather than only its frequent cases.
  • I3. Rare rules are cheap when both ends are given: because matching links known dictionary entries to text, permissive rules for unusual variants rarely cause harm unless such text actually appears.
  • I4. Precision measured as usefulness: a candidate counts as good if it helps the user, not only if it is the single correct choice, which fits candidate-list interfaces.
  • I5. Typed slots in entries: many false hits came from placeholders like "A" or "...", so slot restrictions are a cheap fix.

For Mova

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

  • I1 (frames): idiom and multiword detection could propose all candidate idiom frames first and then discard those whose shape violates explicit, explainable variation rules.
  • I2 (learn): an LLM teacher tier could generate idiom variants as test items, together with negative controls such as literal uses.
  • I3 (world): the compiled idiom lexicon could carry liberal variant licences per entry without much risk of spurious frames.
  • I4 (answer): explanation views could be judged by whether the shown alternatives help a reader, alongside strict accuracy.
  • I5 (tree): idiom entries could be stored as UD subtree templates with typed slots, which would also tolerate long insertions that flat POS patterns miss.

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 6The 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 worksPsychology · 1,552 worksDevelopmental and Educational Psychology · 499 works

🧭 Topics 3Research subjects and languages the work is about.Machine translation · 1,087 worksEnglish · 2,263 worksIdioms & multiword expressions · 217 works

Cite

Cite as Koichi Takeuchi, Takashi Kanehila, Kazuki Hilao, Takeshi Abekawa and Kyo Kageura. 2007. Flexible automatic look-up of English idiom entries in dictionaries. Proceedings of Machine Translation Summit XI: Papers.

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

@inproceedings{takeuchi2007flexible,
  title = {{Flexible automatic look-up of English idiom entries in dictionaries}},
  author = {Koichi Takeuchi, Takashi Kanehila, Kazuki Hilao, Takeshi Abekawa and Kyo Kageura},
  booktitle = {Proceedings of Machine Translation Summit XI: Papers},
  year = {2007},
  url = {https://aclanthology.org/2007.mtsummit-papers.60/}
}

References

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

  1. Abekawa, T. and Kageura, K. (2007). A translation-aid system with a stratified lookup interface. In Proceedings of the 45th Annual Meeting of the Association for Computational Linguistics (to appear). find
  2. Aitchison, J. and Lewis, D. M. (Eds.) (2003). New Media Language. London: Routledge. find
  3. Benson, M. (1985). Collocations and idioms. In Ilson, R. (Ed.) Dictionaries, Lexicography and Language Learning (pp. 61–68). Oxford: Pergamon Press. find
  4. Biber, D. (1999). Longman Grammar of Spoken and Written English. London: Longman. find
  5. Carl, M. and Rascu, E. (2006). A dictionary lookup strategy for translating discontinuous phrases. In The 11th Annual Conference of the European Association for Machine Translation. find
  6. Čermák, F. (2001). Substance of idioms: Perennial problems, lack of data, or theory? International Journal of Lexicography 14(1) (pp. 1–20). ↗
  7. Collins. (2002). Collins COBUILD Phrasal Verbs Dictionary. Glasgow: HarperCollins. find
  8. Excite. (2005). http://www.excite.co.jp/world/. ↗
  9. Fraser, B. (1970). Idioms within a transformational grammar. Foundations of Language 6 (pp. 22–42). find
  10. Fujitsu. (2005). Atlas Personal Translation. find
  11. Jacquemin, C. (2001). Spotting and Discovering Terms through Natural Language Processing. Cambridge, Mass: MIT Press. find
  12. Kageura, K. et. al. (2006). Strategies for enhancing language reference tools to help translators. In Proceedings of the 12th Annual Meeting of the Japan Society of Natural Language Processing (pp. 707–710). find
  13. Kageura, K. and Toyoshima, M. (2006). Analysis of idiom variations in English for the enhanced automatic look-up of idiom entries in dictionaries. Proceedings of the 12th Euralex International Congress (pp. 989–995). find
  14. LogoVista. (2005). LogoVista X PRO 2005. find
  15. McCaleb, J. G. and Iwasaki, M. (2003). All-Purpose Dictionary of English Idioms and Expressions. Tokyo: Asahi Publishing. find
  16. Moon, R. (1998). Fixed Expressions and Idioms in English. Oxford: Clarendon Press. find
  17. Nicolas, T. (1995). Semantics of idiom modification. In Everaert, M. et. al. (Eds.) Idioms: Structural and Psychological Perspectives (pp. 233–252). Hillsdale: Lawrence Erlbaum Associates. find
  18. Oxford. (2001). Oxford Phrasal Verbs: Dictionary for Learners of English. Oxford: Oxford University Press. find
  19. Piao, S. S. L. et. al. (2006). Extracting multiword expressions with a semantic tagger. In Proceedings of ACL2003 Workshop on Multiword Expressions (pp. 143–177). find
  20. Quirk, R. et. al. (1985). A Comprehensive Grammar of the English Language. London: Longman. find
  21. Numberg, G. et. al. (1994). Idioms. In Language 70(3) (pp. 491–538). find
  22. Sanseido. (2004). Grand Concise English-Japanese Dictionary. Tokyo: Sanseido. find
  23. Sharoff, S. et. al. (2006). Using comparable corpora to solve problems difficult for human translators. In Proceedings of the COLING/ACL 2006 (pp. 739–746). find
  24. Sharp. (2004). Hon’yaku Kore Ippom. Similis. http://similis.fr/similis.html. ↗
  25. Smadja, F. (1993). Retrieving collocations from text: Xtract. Computational Linguistics 19(1) (pp. 143–177).
  26. Toshiba Solutions Corporation. (2005). Translation Professional, Version 10. find
  27. Trados. (2006). http://www.trados.com/. ↗
  28. TreeTagger. (2004). http://www.ims.uni-stuttgart.de/projek te/corplex/treetagger/. ↗
  29. Widdows, D. and Dorow, B. (2005). Automatic extraction of idioms using graph analysis and asymmetric lexicosyntactic patterns. In Proceedings of ACL2005 SIGLEX Workshop on Deep Lexical Acquisitions (pp. 48–56). find
  30. Yoshihashi, K. et. al. (2005). Retrieving complex elements from Japanese texts using XPath. In Proceedings of the 11th Annual Meeting of the Japan Society of Natural Language Processing (pp. 257–260). find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/takeuchi-2007-flexible-automatic-look-up-english/