---
title: "An Algorithm for Finding Noun Phrase Correspondences in"
description: "Kupiec addresses the task of automatically pairing noun phrases across an English–French parallel… 12 references, cited by 78 papers here."
authors: ["Julian Kupiec"]
year: 1993
venue: "31st Annual Meeting of the Association for Computational Linguistics"
source: "https://aclanthology.org/P93-1003/"
doi: "10.3115/981574.981577"
licence: "CC BY-NC-SA 3.0 (ACL Anthology, before 2016)"
licence_url: "https://creativecommons.org/licenses/by-nc-sa/3.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/kupiec-1993-algorithm-finding-noun-phrase-correspondences/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/kupiec-1993-algorithm-finding-noun-phrase-correspondences/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [1993](https://md.research.elamur.ai/free/en/papers/1993/) / An Algorithm for Finding Noun Phrase Correspondences in Bilingual Corpora

# An Algorithm for Finding Noun Phrase Correspondences in Bilingual Corpora

**Authors:** [Julian Kupiec](https://md.research.elamur.ai/free/en/authors/julian-kupiec/) · **Year:** 1993 · **Venue:** [31st Annual Meeting of the Association for Computational Linguistics](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 1993](https://md.research.elamur.ai/free/en/venues/acl/1993/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY-NC-SA 3.0 (ACL Anthology, before 2016)](https://creativecommons.org/licenses/by-nc-sa/3.0/)

**Links:** [Publisher's page](https://aclanthology.org/P93-1003/) · [PDF](https://aclanthology.org/P93-1003.pdf) · [DOI 10.3115/981574.981577](https://doi.org/10.3115/981574.981577) · [OpenAlex](https://openalex.org/W2123282296)

**Citations:** 211 citations (OpenAlex, 2026-10-14) · [78](https://md.research.elamur.ai/free/en/papers/kupiec-1993-algorithm-finding-noun-phrase-correspondences/cited-by/) in our library, of the 141,840 papers whose bibliographies we read

## Summary
Kupiec addresses the task of automatically pairing noun phrases across an English–French parallel corpus whose sentences are already aligned, using part of the Canadian Hansards. Each language side goes through an HMM part-of-speech tagger trained on raw Hansard text. Finite-state recognizers, written as regular expressions over tags, then pull out simple noun phrases, allowing embedding only through "of"/"de". These phrases are stored in per-language indexes.

Phrase pairings are learned with EM-style iterative re-estimation, similar to Baum-Welch. At the start, each English phrase in an aligned pair is treated as equally likely to match any French phrase in the partner sentence. Expected counts are summed over the whole corpus, normalised into probabilities, and the loop is repeated.

On a sample of 2,600 alignments the estimates settled after a few iterations, and 90 of the 100 highest-ranked pairs were fully correct. The work shows that light linguistic structure combined with a simple statistical model gives usable phrase-level translation pairs. Such pairs can feed bilingual lexicons, concordances and statistical MT, and the EM approach proved robust and did not need excessive memory.

## Key points
- Data: a 2,600-alignment sample of the aligned Canadian Hansards, yielding about 4,900 distinct English and 5,100 distinct French noun phrases, and 34,000 candidate correspondences after windowing.
- Pipeline: HMM taggers for each language, then regex-based finite-state NP recognizers (no pronouns or digits, leading determiners dropped, singular and plural kept distinct), then phrase indexes, then iterative re-estimation.
- The algorithm is an instance of EM. It works in one direction (English source to French target) but can be reversed by swapping the languages, and it converged in 2–4 iterations.
- Informal evaluation: 90 of the top 100 correspondences were completely correct. Frequent examples include Mr. Speaker ↔ M. le Président and House ↔ Chambre. Rarer correct pairs include acronyms such as DREE ↔ MEER and the long name Atlantic Canada Opportunities Agency.
- Most errors came from the simplified one-phrase-produces-another model rather than from tagging or NP recognition. Free translations such as "toothless tiger" ↔ "tigre de papier" and "roaring rabbit" ↔ "souris qui rugit" broke the model's assumptions.

## Interesting ideas
- **Positional windowing for long sentences:** an NP's relative position in the source sentence is mapped linearly to the target sentence, and only nearby target phrases are treated as candidates. This cheap prior cuts the candidate space and noise, and could help any alignment task on roughly monotone language pairs.
- **Two-stage alignment for word-level links:** first align noun phrases, then treat the resulting phrase pairs as a tiny parallel corpus and rerun the same algorithm with single words as units. Reusing one estimator at two granularities is an easy way to obtain internal word translations, for example for terminology work.
- **Compound sentences for non 1:1 alignments:** when the sentence aligner outputs multi-sentence blocks, they are simply merged into one larger unit with more candidate pairings, so the model itself needs no change.
- **Using cross-lingual repetition to resolve PP attachment:** if a prepositional phrase belongs to a noun phrase, it tends to recur with that NP across occurrences and in both languages. Adverbial PPs do not show this pattern. Thresholds on frequency and correspondence probability can then pick the right bracketing, which could help parsing for any language with a parallel corpus.
- **Rank by probability, temper by frequency:** probabilities order the candidate pairs, while occurrence counts serve as a reliability signal. Singletons can still be resolved correctly when each sentence contains only one NP, which is a useful filter design for building lexicons from low-frequency data.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

## Institutions

- [Palo Alto Research Center](https://md.research.elamur.ai/free/en/labs/palo-alto-research-center/) — Research institute, Palo Alto, United States · 11 works · 5,515 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,086 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,889 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 9,296 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 9,206 works

## Cite

Cite as: Julian Kupiec. 1993. An Algorithm for Finding Noun Phrase Correspondences in Bilingual Corpora. 31st Annual Meeting of the Association for Computational Linguistics.

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

```bibtex
@inproceedings{kupiec1993algorithm,
  title = {{An Algorithm for Finding Noun Phrase Correspondences in Bilingual Corpora}},
  author = {Julian Kupiec},
  booktitle = {31st Annual Meeting of the Association for Computational Linguistics},
  year = {1993},
  doi = {10.3115/981574.981577},
  url = {https://doi.org/10.3115/981574.981577}
}
```

- [← previous Aligning Sentences in Bilingual Corpora Using Lexical Information](https://md.research.elamur.ai/free/en/papers/chen-1993-aligning-sentences-bilingual-corpora-lexical/)
- [next → Two-level Description of Turkish Morphology](https://md.research.elamur.ai/free/en/papers/oflazer-1993-two-level-description-turkish-morphology/)
- [↑ back to papers of 1993](https://md.research.elamur.ai/free/en/papers/1993/)

## References

12 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. P. F. Brown, J. C. Lai, and R. L. Mercer. [Aligning sentences in parallel corpora](https://md.research.elamur.ai/free/en/papers/brown-1991-aligning-sentences-parallel-corpora/). In Proceedings of the 29th Annual Meeting of the Association of Computational Linguistics, pages 169-176, Berkeley, CA., June 1991.
2. [P. F. Brown, S. A. Della Pietra, V. J. Della Pietra, and R. L. Mercer. Word sense disambiguation using statistical methods. In Proceedings of the 29th Annual Meeting of the Association of Computational Linguistics, pages 264-270, Berkeley, CA., June 1991.](https://md.research.elamur.ai/free/en/papers/brown-1991-word-sense-disambiguation-statistical-methods/)
3. P. F. Brown, S. A. Della Pietra, V. J. Della Pietra, J. D. Lafferty, and R. L. Mercer. Analysis, statistical transfer, and synthesis in machine translation. In Proceedings of the Fourth International Conference on Theoretical and Methodological Issues in Machine Translation, pages 83-100, Montreal,… [find](https://scholar.google.com/scholar?q=P%2E+F%2E+Brown%2C+S%2E+A%2E+Della+Pietra%2C+V%2E+J%2E+Della+Pietra%2C+J%2E+D%2E+Lafferty%2C+and+R%2E+L%2E+Mercer%2E+Analysis%2C+statistical+transfer%2C+and+synthesis+in+machine+translation%2E+In+Proceedings+of+the+Fourth+International+)
4. K. W. Church and W. A. Gale. Concordances for parallel text. In Proceedings of the Seventh Annual Conference of the UW Center for the New OED and Text Research, pages 40-62, September 1991. [find](https://scholar.google.com/scholar?q=K%2E+W%2E+Church+and+W%2E+A%2E+Gale%2E+Concordances+for+parallel+text%2E+In+Proceedings+of+the+Seventh+Annual+Conference+of+the+UW+Center+for+the+New+OED+and+Text+Research%2C+pages+40%2D62%2C+September+1991%2E)
5. D. Cutting, J. Kupiec, J. Pedersen, and P. Sibun. [A practical part-of-speech tagger](https://md.research.elamur.ai/free/en/papers/cutting-1992-practical-part-speech-tagger/). In Proceedings of the Third Conference on Applied Natural Language Processing, Trento, Italy, April 1992. ACL.
6. I. Dagan, A. Itai, and U. Schwall. [Two languages are more informative than one](https://md.research.elamur.ai/free/en/papers/dagan-1991-two-languages-are-more-informative/). In Proceedings of the 29th Annual Meeting of the Association of Computational Linguistics, pages 130-137, Berkeley, CA., June 1991.
7. A.P. Dempster, N.M. Laird, and D.B. Rubin. Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society, B39:1-38, 1977. [find](https://scholar.google.com/scholar?q=A%2EP%2E+Dempster%2C+N%2EM%2E+Laird%2C+and+D%2EB%2E+Rubin%2E+Maximum+likelihood+from+incomplete+data+via+the+EM+algorithm%2E+Journal+of+the+Royal+Statistical+Society%2C+B39%3A1%2D38%2C+1977%2E)
8. W. A. Gale and K. W. Church. Identifying word correspondences in parallel texts. In Proceedings of the Fourth DARPA Speech and Natural Language Workshop, pages 152-157, Pacific Grove, CA., February 1991. Morgan Kaufmann. [find](https://scholar.google.com/scholar?q=W%2E+A%2E+Gale+and+K%2E+W%2E+Church%2E+Identifying+word+correspondences+in+parallel+texts%2E+In+Proceedings+of+the+Fourth+DARPA+Speech+and+Natural+Language+Workshop%2C+pages+152%2D157%2C+Pacific+Grove%2C+CA%2E%2C+February+19)
9. W. A. Gale and K. W. Church. [A program for aligning sentences in bilingual corpora](https://md.research.elamur.ai/free/en/papers/gale-1993-program-aligning-sentences-bilingual-corpora/). In Proceedings of the 29th Annual Meeting of the Association of Computational Linguistics, pages 177-184, Berkeley, CA., June 1991.
10. M. Kay and M. RSscheisen. Text-translation alignment. Technical Report P90-00143, Xerox Palo Alto Research Center, 3333 Coyote Hill Rd., Palo Alto, CA 94304, June 1988. [find](https://scholar.google.com/scholar?q=M%2E+Kay+and+M%2E+RSscheisen%2E+Text%2Dtranslation+alignment%2E+Technical+Report+P90%2D00143%2C+Xerox+Palo+Alto+Research+Center%2C+3333+Coyote+Hill+Rd%2E%2C+Palo+Alto%2C+CA+94304%2C+June+1988%2E)
11. J. M. Kupiec. Robust part-of-speech tagging using a hidden markov model. Computer Speech and Language, 6:225-242, 1992. [find](https://scholar.google.com/scholar?q=J%2E+M%2E+Kupiec%2E+Robust+part%2Dof%2Dspeech+tagging+using+a+hidden+markov+model%2E+Computer+Speech+and+Language%2C+6%3A225%2D242%2C+1992%2E)
12. F. Smadja. How to compile a bilingual collocational lexicon automatically. In C. Weir, editor, Proceedings of the AAAI-92 Workshop on Statistically-Based NLP Techniques, San Jose, CA, July 1992. [find](https://scholar.google.com/scholar?q=F%2E+Smadja%2E+How+to+compile+a+bilingual+collocational+lexicon+automatically%2E+In+C%2E+Weir%2C+editor%2C+Proceedings+of+the+AAAI%2D92+Workshop+on+Statistically%2DBased+NLP+Techniques%2C+San+Jose%2C+CA%2C+July+1992%2E)

*Updated 24 October 2026*