---
title: "Phrasal Cohesion and Statistical Machine Translation — Fox 2002"
description: "This paper tests an assumption behind syntax-aware statistical machine translation: that phrases stay… 10 references, cited by 86 papers here."
authors: ["Heidi Fox"]
year: 2002
venue: "Proceedings of the 2002 Conference on Empirical Methods in Natural Language Processing (EMNLP 2002)"
source: "https://aclanthology.org/W02-1039/"
doi: "10.3115/1118693.1118732"
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/fox-2002-phrasal-cohesion-statistical-machine-translation/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/fox-2002-phrasal-cohesion-statistical-machine-translation/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2002](https://md.research.elamur.ai/free/en/papers/2002/) / Phrasal Cohesion and Statistical Machine Translation

# Phrasal Cohesion and Statistical Machine Translation

**Authors:** Heidi Fox · **Year:** 2002 · **Venue:** [Proceedings of the 2002 Conference on Empirical Methods in Natural Language Processing (EMNLP 2002)](https://md.research.elamur.ai/free/en/venues/emnlp/) ([EMNLP 2002](https://md.research.elamur.ai/free/en/venues/emnlp/2002/)) · **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/W02-1039/) · [PDF](https://aclanthology.org/W02-1039.pdf) · [DOI 10.3115/1118693.1118732](https://doi.org/10.3115/1118693.1118732) · [OpenAlex](https://openalex.org/W2160382364)

**Citations:** 164 citations (OpenAlex, 2026-10-14) · [86](https://md.research.elamur.ai/free/en/papers/fox-2002-phrasal-cohesion-statistical-machine-translation/cited-by/) in our library, of the 141,840 papers whose bibliographies we read

## Summary
This paper tests an assumption behind syntax-aware statistical machine translation: that phrases stay together, or "cohere", when moved from one language to another. Fox measures this empirically for English→French. The data are 500 Canadian Hansard sentence pairs with manual word alignments (Och and Ney, 2000), and the English side is parsed with the Charniak (2000) parser trained on the Penn Treebank. Each English constituent is projected onto a French span. When sibling spans overlap, this counts as a "crossing". Crossings are tallied separately for head–modifier and modifier–modifier pairs, with and without a filter that ignores phrases translated as a whole unit. Crossings are rare with sure alignments but much more frequent once possible alignments are added. Even so, many crossings come from free rewording or parser errors, and most of the rest come from a small set of recurring constructions. Of three tree representations, dependency structures cohere best. The study supports phrase-movement reordering and points to where it will fail.

## Key points
- **Data and conditions:** 500 manually aligned Hansard sentence pairs with sure (S) and possible (P) links. Three alignment conditions were tested: S only, P only, and S with P as a fallback (S∪P).
- **Average crossings per sentence (no filter):**
  - With S links: 0.236 head crossings and 0.056 modifier crossings.
  - With S∪P links: these rise to 4.790 and 0.880.
- **Effect of the phrasal-translation filter (S∪P):** head crossings drop to 2.772 and modifier crossings to 0.516. Phrasal translations account for nearly half of all crossings.
- **Comparison of representations (S∪P, filter on):**

  | Representation | Avg. head crossings | Head crossings per chance | Modifier crossings per chance |
  |---|---|---|---|
  | Treebank trees (baseline) | 2.772 | 18.61% | 8.47% |
  | Flattened verb phrases | 2.252 | 15.12% | 10.59% |
  | Dependencies | 1.88 | 12.62% | 9.22% |

- **Overall result:** dependency structures show the best overall cohesion. The author treats the S∪P figures as an upper bound, because the heuristic misses some phrasal translations and P links merge conflicting annotator choices.

## Interesting ideas
- **Span-overlap ("crossing") counting:** this is a simple, model-free way to check whether a syntactic representation is compatible with reordering by moving whole subtrees. It could be reused to compare treebanks, annotation schemes or parsers for any language pair before building a system.
- **Normalizing by opportunities:** reporting crossings as a percentage of possible tests, not just raw counts, removes a bias. Different structures offer different numbers of head and modifier pairs, so this is a fairer way to compare representations.
- **Filtering unit translations:** whole-chunk translations such as idioms or free renderings are detected and excluded so they do not inflate structural error counts. This is a useful step whenever alignment-based metrics mix lexical and structural divergence.
- **Small pilot annotation:** since a few constructions cause most breakdowns, manually aligning a small sample can reveal the main trouble spots for a new language pair. Those findings can then guide corpus filtering or targeted model fixes, which is cheap diagnostics for low-resource pairs such as those involving Ukrainian.
- **Probing representation choices:** comparing Treebank trees, flattened verb phrases and dependencies on one fixed metric shows that the choice of representation has measurable trade-offs. Flattening reduces head crossings but adds modifier crossings.

*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

- [Brown University](https://md.research.elamur.ai/free/en/universities/brown-university/) — University, Providence, United States · 58 works · 25,310 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,085 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,888 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

🧭 Topics (1): [🔁 Machine translation](https://md.research.elamur.ai/free/en/topics/machine-translation/) · 958 works

## Cite

Cite as: Heidi Fox. 2002. Phrasal Cohesion and Statistical Machine Translation. Proceedings of the 2002 Conference on Empirical Methods in Natural Language Processing (EMNLP 2002).

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

```bibtex
@inproceedings{fox2002phrasal,
  title = {{Phrasal Cohesion and Statistical Machine Translation}},
  author = {Heidi Fox},
  booktitle = {Proceedings of the 2002 Conference on Empirical Methods in Natural Language Processing (EMNLP 2002)},
  year = {2002},
  doi = {10.3115/1118693.1118732},
  url = {https://doi.org/10.3115/1118693.1118732}
}
```

- [← previous Pronunciation Modeling for Improved Spelling Correction](https://md.research.elamur.ai/free/en/papers/toutanova-2002-pronunciation-modeling-improved-spelling-correction/)
- [next → Entropy Rate Constancy in Text](https://md.research.elamur.ai/free/en/papers/genzel-2002-entropy-rate-constancy-text/)
- [↑ back to papers of 2002](https://md.research.elamur.ai/free/en/papers/2002/)

## References

10 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Hiyan Alshawi, Srinivas Bangalore, and Shona Douglas. 2000. Learning dependency translation models as collections of finite-state head transducers. Computational Linguistics, 26(1):45–60, March. [find](https://scholar.google.com/scholar?q=Hiyan+Alshawi%2C+Srinivas+Bangalore%2C+and+Shona+Douglas%2E+2000%2E+Learning+dependency+translation+models+as+collections+of+finite%2Dstate+head+transducers%2E+Computational+Linguistics%2C+26%281%29%3A45%E2%80%9360%2C+March%2E)
2. Peter F. Brown, Stephen A. Della Pietra, Vincent J. Della Pietra, and Robert L. Mercer. 1993. The mathematics of machine translation: Parameter estimation. Computational Linguistics, 19(2):263–311, June. [find](https://scholar.google.com/scholar?q=Peter+F%2E+Brown%2C+Stephen+A%2E+Della+Pietra%2C+Vincent+J%2E+Della+Pietra%2C+and+Robert+L%2E+Mercer%2E+1993%2E+The+mathematics+of+machine+translation%3A+Parameter+estimation%2E+Computational+Linguistics%2C+19%282%29%3A263%E2%80%93311%2C+Ju)
3. Eugene Charniak. 2000. [A maximum-entropy-inspired parser](https://md.research.elamur.ai/free/en/papers/charniak-2000-maximum-entropy-inspired-parser/). In Proceedings of the 1st Meeting of the North American Chapter of the Association for Computational Linguistics.
4. Douglas Jones and Rick Havrilla. 1998. Twisted pair grammar: Support for rapid development of machine translation for low density languages. In Proceedings of the Conference of the Association for Machine Translation in the Americas. [find](https://scholar.google.com/scholar?q=Douglas+Jones+and+Rick+Havrilla%2E+1998%2E+Twisted+pair+grammar%3A+Support+for+rapid+development+of+machine+translation+for+low+density+languages%2E+In+Proceedings+of+the+Conference+of+the+Association+for+Mac)
5. Mitchell P. Marcus, Beatrice Santorini, and Mary Ann Marcinkiewicz. 1993. [Building a large annotated corpus of English: The Penn Treebank](https://md.research.elamur.ai/free/en/papers/marcus-1993-building-large-annotated-corpus-english/). Computational Linguistics, 13(2):313–330, June.
6. Franz Josef Och and Hermann Ney. 2000. [Improved statistical alignment models](https://md.research.elamur.ai/free/en/papers/och-2000-improved-statistical-alignment-models/). In Proceedings of the 38th Annual Meeting of the Association for Computational Linguistics, pages 440–447.
7. Ye-Yi Wang. 1998. Grammar Inference and Statistical Machine Translation. Ph.D. thesis, Carnegie Mellon University. [find](https://scholar.google.com/scholar?q=Ye%2DYi+Wang%2E+1998%2E+Grammar+Inference+and+Statistical+Machine+Translation%2E+Ph%2ED%2E+thesis%2C+Carnegie+Mellon+University%2E)
8. Dekai Wu. 1997. [Stochastic inversion transduction grammars and bilingual parsing of parallel corpora](https://md.research.elamur.ai/free/en/papers/wu-1997-stochastic-inversion-transduction-grammars-bilingual/). Computational Linguistics, 23(3):377–403, September.
9. Kenji Yamada and Kevin Knight. 2001. [A syntax-based statistical translation model](https://md.research.elamur.ai/free/en/papers/yamada-2001-syntax-based-statistical-translation-model/). In Proceedings of the 39th Annual Meeting of the Association for Computational Linguistics.
10. David Yarowsky, Grace Ngai, and Richard Wicentowski. 2001. [Inducing multilingual text analysis tools via robust projection across aligned corpora](https://md.research.elamur.ai/free/en/papers/yarowsky-2001-inducing-multilingual-text-analysis-tools/). In Proceedings of the ARPA Human Language Technology Workshop, pages 109–116.

*Updated 24 October 2026*