Elamur Research⌕
☰
home / papers / 1999 / Automatic Identification of Word Translations from Unrelated English and German Corpora

Automatic Identification of Word Translations from Unrelated English and German Corpora

Summary

The paper addresses how to find word translations when no parallel text exists, using only two monolingual corpora that are largely unrelated in content. The usual alignment clues (word/sentence order, frequency correlation, cognates) mostly fail in this setting. The author therefore relies on one assumption: if two words co-occur unusually often in one language, their translations should also co-occur unusually often in the other.

The method starts from a small German–English seed dictionary. For each German word, it builds a co-occurrence vector over the seed words and maps the known entries into English. It then compares this vector with the vectors of all English words built from the English corpus, and the closest English vector is taken as the translation.

The data are the Frankfurter Allgemeine Zeitung and the Guardian. Both corpora were lemmatized and stripped of function words. Co-occurrences are counted separately for each window position, and raw counts are replaced by association strengths.

On 100 test words, about 72% were translated correctly, compared with roughly 30% in earlier non-parallel work. This shows that cheap monolingual text can be used to extend bilingual dictionaries.

Key points

  • Corpora: 135 million words of Frankfurter Allgemeine Zeitung (1993–1996) and 163 million words of the Guardian (1990–1994). The papers differ in orientation and only partly overlap in time, so the corpora are treated as unrelated.
  • Seed lexicon: derived from the Collins Gem German Dictionary (about 22,300 entries). Removing multi-word entries left 16,380, and all entries for the test words were then removed as well.
  • Test set: 100 German stimulus words from Russell's (1970) word association norms, with their English translations as the reference.
  • Preprocessing: function words were removed (lists of about 600 German and about 200 English words). Lemmatization used only a morphological lexicon, without context. The window size was 3, with a separate vector for each position, so word order is preserved.
  • Results: about 72% correct, versus about 30% for prior non-parallel methods and up to 99% reported for parallel texts. Ambiguity explains some failures. For example, the expected translation of Kohl ("cabbage") is ranked 17,074th, and the top candidates are politicians such as Major, Thatcher and Gorbachev.

Interesting ideas

  • I1. A seed dictionary makes the search cheap. Assuming a small starting lexicon turns an infeasible search over matrix permutations into simple vector comparisons. This is a practical way to bootstrap lexicons for any low-resource language pair where only a basic dictionary exists, such as Ukrainian paired with a smaller language.
  • I2. Position-specific co-occurrence vectors. Instead of summing counts over a window, the method concatenates one vector per relative position. This keeps rough word-order information at low cost and could help in other distributional similarity tasks.
  • I3. Symmetric preprocessing across languages. The author deliberately skipped a context-sensitive German lemmatizer because no English equivalent was available, so both sides were treated alike. Keeping pipelines comparable matters whenever two languages' representations are compared directly.
  • I4. Leakage-free evaluation design. The test words were fixed in advance and their entries removed from the seed lexicon, so the method could never see the answers. The author also reports the rank of the expected translation rather than just hit or miss. Both habits make evaluation more honest and error analysis more informative.
  • I5. Precision-only scoring, with a stated reason. Recall was not measured because it depends heavily on how large the reference dictionary is. Being explicit about which metrics are meaningful for a given gold standard is worth copying.

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: Natural language processing

Cited by

203 papers of our library cite it (of the 140884 whose bibliographies we read); 6 of them have a page here, 197 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.3115/1034678.1034756

@inproceedings{rapp1999automatic,
  title = {{Automatic Identification of Word Translations from Unrelated English and German Corpora}},
  author = {Rapp, Reinhard},
  booktitle = {Proceedings of the 37th Annual Meeting of the Association for Computational Linguistics},
  year = {1999},
  doi = {10.3115/1034678.1034756},
  url = {https://doi.org/10.3115/1034678.1034756}
}

References

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

  1. Armstrong, S.; Kempen, M.; Petitpierre, D.; Rapp, R.; Thompson, H. (1998). Multilingual Corpora for Cooperation. Proceedings of the 1st International Conference on Linguistic Resources and Evaluation (LREC), Granada, Vol. 2, 975-980. find
  2. Brown, P.; Cocke, J.; Della Pietra, S. A.; Della Pietra, V. J.; Jelinek, F.; Lafferty, J. D.; Mercer, R. L.; Rossin, P. S. (1990). A statistical approach to machine translation. Computational Linguistics, 16(2), 79-85. find
  3. Church, K. W.; Hanks, P. (1989). Word association norms, mutual information, and lexicography. In: Proceedings of the 27th Annual Meeting of the Association for Computational Linguistics. Vancouver, British Columbia, 76-83. find our card
  4. Dunning, T. (1993). Accurate methods for the statistics of surprise and coincidence. Computational Linguistics, 19(1), 61-74. find our card
  5. Fung, P. (1995). Compiling bilingual lexicon entries from a non-parallel English-Chinese corpus. Proceedings of the 3rd Annual Workshop on Very Large Corpora, Boston, Massachusetts, 173-183. find
  6. Fung, P.; McKeown, K. (1997). Finding terminology translations from non-parallel corpora. Proceedings of the 5th Annual Workshop on Very Large Corpora, Hong Kong, 192-202. find
  7. Fung, P.; Yee, L. Y. (1998). An IR approach for translating new words from nonparallel, comparable texts. In: Proceedings of COLING-ACL 1998, Montreal, Vol. 1,414-420. find our card
  8. Gale, W. A.; Church, K. W. (1993). A program for aligning sentences in bilingual corpora. Computational Linguistics, 19(3), 75-102. find our card
  9. Grefenstette, G. (1993). Evaluation techniques for automatic semantic extraction: comparing syntactic and window based approaches. In: Proceedings of the Workshop on Acquisition of Lexical Knowledge from Text, Columbus, Ohio. find
  10. Grefenstette, G. (1994). Explorations in Automatic Thesaurus Discovery. Dordrecht: Kluwer. find
  11. Jones, W. P.; Furnas, G. W. (1987). Pictures of relevance: a geometric analysis of similarity measures. Journal of the American Society for Information Science, 38(6), 420-442. find
  12. Kay, M.; Rfscheisen, M. (1993). Text-Translation Alignment. Computational Linguistics, 19(1), 121-142. find our card
  13. Landauer, T. K.; Dumais, S. T. (1997). A solution to Plato's problem: the latent semantic analysis theory of acquisition, induction, and representation of knowledge. Psychological Review, 104(2), 211-240. find
  14. Langlais, P.; Simard, M.; V6ronis, J. (1998). Methods and practical issues in evaluating alignment techniques. In: Proceedings of COLING-ACL 1998, Montreal, Vol. l, 711-717. find
  15. Lezius, W.; Rapp, R.; Wettler, M. (1998). A freely available morphology system, part-of-speech tagger, and context-sensitive lemmatizer for German. In: Proceedings of COLING-ACL 1998, Montreal, Vol. 2, 743-748. find
  16. Lin, D. (1998). Automatic Retrieval and Clustering of Similar Words. In: Proceedings of COLING-ACL 1998, Montreal, Vol. 2, 768-773. find our card
  17. Rapp, R. (1995). Identifying word translations in non-parallel texts. In: Proceedings of the 33rd Meeting of the Association for Computational Linguistics. Cambridge, Massachusetts, 320-322. find our card
  18. Rapp, R. (1996). Die Berechnung von Assoziationen. Hildesheim: Olms. find
  19. Ruge, G. (1995). Human memory models and term association. Proceedings of the ACM SIGIR Conference, Seattle, 219-227. find
  20. Russell, W. A. (1970). The complete German language norms for responses to 100 words from the Kent-Rosanoff word association test. In: L. Postman, G. Keppel (eds.): Norms of Word Association. New York: Academic Press, 53-94. find
  21. Salton, G.; McGill, M. (1983). Introduction to Modem Information Retrieval. New York: McGraw-Hill. find
  22. Schiitze, H. (1993). Part-of-speech induction from scratch. In: Proceedings of the 31st Annual Meeting of the Association for Computational Linguistics, Columbus, Ohio, 251-258. find
  23. Wettler, M.; Rapp, R. (1993). Computation of word associations based on the co-occurrences of words in large corpora. In: Proceedings of the 1st Workshop on Very Large Corpora: Columbus, Ohio, 84-93. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/rapp-1999-automatic-identification-word-translations-unrelated/