Elamur Research
☰
home / papers / 1993 / A Program for Aligning Sentences in Bilingual Corpora

A Program for Aligning Sentences in Bilingual Corpora

Written for the Research library (claude-opus-5-5) from the paper text, in our own words: the license "CC BY-NC-SA 3.0 (ACL Anthology, before 2016)" does not allow redistributing the text itself. Original — aclanthology.org/J93-1004.pdf

Summary

Gale and Church tackle sentence alignment: deciding which sentences in a text correspond to which sentences in its translation. This is a prerequisite for statistical machine translation and bilingual lexicography. Earlier systems were mostly unpublished, slow, or poorly evaluated. The authors present a program, align, that uses no lexical information. It relies only on the observation that sentence lengths, counted in characters, are strongly correlated across translations.

Each candidate pairing is scored probabilistically from the normalized length difference and its variance, including 1-1, 1-2, 2-1 and 2-2 matches. Dynamic programming then selects the most likely overall alignment. Paragraphs are aligned first, then sentences within each paragraph.

On a trilingual UBS corpus of banking reports in English, French and German, about 4% of alignments were wrong. Keeping only the highest-scoring alignments yields a much cleaner subset. The authors also aligned roughly 90 million words of Canadian Hansards for release through ACL/DCI. The method's speed, simplicity and openness made large parallel corpora practical to use.

Key points

  • The evaluation data were 15 UBS economic reports, with 14,680 words, 725 sentences and 188 paragraphs on the English side. Across English-French and English-German, the overall error rate was 4.2% on 1316 alignments, with English-French producing more errors.
  • Restricting output to the best-scoring 80% of alignments cut the error rate to 0.7%.
  • Counting length in characters worked better than counting words. Word-based lengths raised the error rate from 4.2% to 6.5%. Paragraph lengths and their translations correlated at 0.991.
  • The same language-independent parameters were used for both language pairs. Fitting pair-specific values left the total error count unchanged.
  • The aligned Canadian Hansards (about 90M words) were released via ACL/DCI, and the paper includes C code for the program's core. Later extensions added hard and soft delimiters, and the authors suggest lexical cues as a future fallback.

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: Machine translation · Natural language processing

Cited by

339 papers of our library cite it (of the 140673 whose bibliographies we read); 8 of them have a page here, 331 not yet.

Cite

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

@article{gale1993program,
  title = {{A Program for Aligning Sentences in Bilingual Corpora}},
  author = {Gale, William A. and Church, Kenneth W.},
  journal = {Computational Linguistics},
  year = {1993},
  url = {https://aclanthology.org/J93-1004/}
}

References

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

  1. Abramowitz, M., and Stegun, I. (1964). Handbook of Mathematical Functions. US Government Printing Office. find
  2. Brown, P.; Cocke, J.; Della Pietra, S.; Della Pietra, V.; Jelinek, F.; Mercer, R.; and Roossin, P. (1988a). "A statistical approach to French/English translation." In Proceedings, RIA088 Conference. Cambridge, MA. find
  3. Brown, P.; Cocke, J.; Della Pietra, S.; Della Pietra, V.; Jelinek, F.; Mercer, R.; and Roossin, P. (1988b). "A statistical approach to language translation." In Proceedings, 13th International Conference on Computational Linguistics (COLING-88). Budapest, Hungary. find
  4. Brown, P.; Cocke, J.; Della Pietra, S.; Della Pietra, V.; Jelinek, F.; Lafferty, J.; Mercer, R.; and Roossin, P. (1990). "A statistical approach to machine translation." Computational Linguistics, 16, 79-85. find
  5. Brown, P.; Lai, J.; and Mercer, R. (1991). "Aligning sentences in parallel corpora." In Proceedings, 47th Annual Meeting of the Association for Computational Linguistics. find
  6. Catizone, R.; Russell, G.; and Warwick, S. (in press). "Deriving translation data from bilingual texts." In Lexical Acquisition: Using on-line Resources to Build a Lexicon, edited by Zernik. Lawrence Erlbaum. find
  7. Church, K. (1988). "A stochastic parts program and noun phrase parser for unrestricted text." In Proceedings, Second Conference on Applied Natural Language Processing. Austin, TX. find
  8. Hosmer, D., and Lemeshow, S. (1989). Applied Logistic Regression. Wiley. find
  9. Klavans, J., and Tzoukermann, E. (1990). "The BICORD system." In Proceedings, 15th International Conference on Computational Linguistics (COLING-90), 174-179. find
  10. Kay, M., and R6scheisen, M. (1988). "Text-translation alignment." Xerox Palo Alto Research Center. find
  11. Kruskal, J., and Liberman, M. (1983). "The symmetric time-warping problem: From continuous to discrete." In Time Warps, String Edits, and Macro Molecules: The Theory and Practice of Sequence Comparison, edited by D. Sankoff and J. Kruskal. Addison-Wesley. find
  12. Liberman, M., and Church, K. (in press). "Text analysis and word pronunciation in text-to-speech synthesis." In Advances in Speech Signal Processing, edited by S. Furui and M. Sondhi. find
  13. Sankoff, D., and Kruskal, J. (1983). Time Warps, String Edits, and Macromolecules: The Theory and Practice of Sequence Comparison. Addison-Wesley. find