---
title: "Improving Bitext Word Alignments via Syntax-based Reordering"
description: "Statistical word aligners such as the IBM models handle local word-order shifts well but struggle when… 11 references, cited by 2 papers here."
authors: ["Elliott Franco Drabek", "David Yarowsky"]
year: 2004
venue: "Proceedings of the ACL Interactive Poster and Demonstration Sessions"
source: "https://aclanthology.org/P04-3014/"
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/drabek-2004-improving-bitext-word-alignments-syntax/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/drabek-2004-improving-bitext-word-alignments-syntax/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2004](https://md.research.elamur.ai/free/en/papers/2004/page/2/) / Improving Bitext Word Alignments via Syntax-based Reordering of English

# Improving Bitext Word Alignments via Syntax-based Reordering of English

**Authors:** [Elliott Franco Drabek](https://md.research.elamur.ai/free/en/authors/elliott-franco-drabek/) and [David Yarowsky](https://md.research.elamur.ai/free/en/authors/david-yarowsky/) · **Year:** 2004 · **Venue:** [Proceedings of the ACL Interactive Poster and Demonstration Sessions](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2004](https://md.research.elamur.ai/free/en/venues/acl/2004/)) · **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/P04-3014/) · [PDF](https://aclanthology.org/P04-3014.pdf) · [OpenAlex](https://openalex.org/W2097377186)

**Citations:** 5 citations (OpenAlex, 2026-10-14) · 2 in our library, of the 149,439 papers whose bibliographies we read

## Summary
Statistical word aligners such as the IBM models handle local word-order shifts well but struggle when two languages differ in long-range constituent order. Parsers for low-resource languages are usually unavailable. Drabek and Yarowsky avoid needing one. They parse only the English side with the Collins parser and apply hand-written, language-specific rules that permute tree children so English mimics the other language's basic order (verb–object, adposition–object, noun–modifier). The result is a reordered "English′". They align it with the foreign text using unmodified GIZA++, then map the links back to the original English words through recorded traces. The required knowledge is coarse and can be taken from a brief grammar sketch. Experiments cover Hindi, Korean, Chinese and Romanian, with varying amounts of bitext and a dictionary-only setting. Hindi and Korean, both verb-final, gain consistently. Chinese and Romanian, already close to English order, show no notable change. Because the method only pre- and post-processes data, it can wrap any existing aligner.

## Key points
- Pipeline: the Collins parser runs on English, rule-based reordering produces English′ with traces, GIZA++ aligns it, and the traces project the alignments back to the original English.
- Table 1 records each language's coarse word order. Hindi and Korean are OV with postpositions. Chinese has both prepositions and postpositions, so it gets extra rules: a bracketing postposition is added, and non-NP verb modifiers move left of the verb.
- Test sets: 46 Hindi, 100 Korean, 88 Chinese and 248 Romanian sentence pairs with gold alignments. Scoring used Mihalcea and Pedersen's wa_eval_align.pl.
- Reordering raised the correlation (r²) between English and foreign normalized word positions:
  - Korean: 10.2 → 31.6
  - Hindi: 54.1 → 60.1
  - Chinese: 60.2 → 63.7
  - Romanian: roughly unchanged (81.1 → 80.6)
- Evaluation used learning curves over growing training bitext, starting at 1000 sentences, plus a "dictionary only" condition with no parallel text for Hindi, Korean and Chinese. Hindi and Korean improved in every condition.

## Interesting ideas
- Instead of building a more complex alignment model, the authors rewrite the input so a simple, robust model works better, then undo the rewrite using stored traces; this wrapper trick reuses tools without touching them.
- Coarse typological facts (OV/VO, pre- vs postpositions, adjective and relative-clause position) are enough to drive useful tree transformations, so a few parameters stand in for a full grammar.
- Measuring the correlation of normalized word positions across aligned pairs gives a cheap, interpretable index of how much reordering separates two texts, and shows whether a transformation actually helped.
- The approach is tested where it should help (head-final languages) and where it should not (Romanian, Chinese), and the near-zero effect on the second group acts as a natural control.
- Running the system with only a bilingual word list isolates how much the reordering rules contribute on their own, apart from what is learned from data.

## For Mova

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

- **I1** (back): Translation through the meta-world could linearize English trees into target-language order using reversible, traced permutations, so every reordering step can be explained and undone.
- **I2** (tree): UD dialect conversion and cross-lingual treebank work could be driven by a small table of word-order parameters applied to head–dependent orderings, rather than per-language code.
- **I3** (learn): A position-correlation score could serve as a sanity check when comparing witness parsers' outputs or silver data against references, flagging systematic order mismatches.
- **I4** (learn): When evaluating repair rules, include inputs where a rule should change nothing, and treat any shift there as a negative-control failure.
- **I5** (learn): Ablations that remove trained components and keep only compiled knowledge would show how much explainable rule knowledge carries versus learned statistics.

*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

- [Johns Hopkins University](https://md.research.elamur.ai/free/en/universities/johns-hopkins-university/) — University, Baltimore, United States · 240 works · 96,487 citations

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (2): [English](https://md.research.elamur.ai/free/en/topics/english/) · 2,263 works · [📐 Syntax & grammar formalisms](https://md.research.elamur.ai/free/en/topics/syntax-grammar/) · 214 works

## Cite

Cite as: Elliott Franco Drabek and David Yarowsky. 2004. Improving Bitext Word Alignments via Syntax-based Reordering of English. Proceedings of the ACL Interactive Poster and Demonstration Sessions.

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

```bibtex
@inproceedings{drabek2004improving,
  title = {{Improving Bitext Word Alignments via Syntax-based Reordering of English}},
  author = {Elliott Franco Drabek and David Yarowsky},
  booktitle = {Proceedings of the ACL Interactive Poster and Demonstration Sessions},
  year = {2004},
  doi = {10.3115/1219044.1219058},
  url = {https://doi.org/10.3115/1219044.1219058}
}
```

- [← previous An Algorithm for Resolving Individual and Abstract Anaphora in Danish Texts and Dialogues](https://md.research.elamur.ai/free/en/papers/navarretta-2004-algorithm-resolving-individual-abstract-anaphora/)
- [next → NLML — a Markup Language to Describe the Unlimited English Grammar](https://md.research.elamur.ai/free/en/papers/jia-2004-nlml-markup-language-describe-unlimited/)
- [↑ back to papers of 2004](https://md.research.elamur.ai/free/en/papers/2004/page/2/)

## References

11 references, 4 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, S. A. Della Pietra, V. J. Della Pietra, and R. L. Mercer. 1993. The mathematics of statistical machine translation: Parameter estimation. Computational Linguistics, 19(2):263–311. [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+and+R%2E+L%2E+Mercer%2E+1993%2E+The+mathematics+of+statistical+machine+translation%3A+Parameter+estimation%2E+Computational+Linguistics%2C+19%282%29%3A263%E2%80%93311%2E)
2. M. Collins. 1999. Head-Driven Statistical Models for Natural Language Parsing. Ph.D. thesis, University of Pennsylvania. [find](https://scholar.google.com/scholar?q=M%2E+Collins%2E+1999%2E+Head%2DDriven+Statistical+Models+for+Natural+Language+Parsing%2E+Ph%2ED%2E+thesis%2C+University+of+Pennsylvania%2E)
3. B. J. Dorr, L. Pearl, R. Hwa, and N. Habash. 2002. DUSTer: A method for unraveling cross-language divergences for statistical word-level alignment. In Proceedings of AMTA-02, pages 31–43. [find](https://scholar.google.com/scholar?q=B%2E+J%2E+Dorr%2C+L%2E+Pearl%2C+R%2E+Hwa%2C+and+N%2E+Habash%2E+2002%2E+DUSTer%3A+A+method+for+unraveling+cross%2Dlanguage+divergences+for+statistical+word%2Dlevel+alignment%2E+In+Proceedings+of+AMTA%2D02%2C+pages+31%E2%80%9343%2E)
4. A. Lopez, M. Nosal, R. Hwa, and P. Resnik. 2002. Word-level alignment for multilingual resource acquisition. In Proceedings of the LREC-02 Workshop on Linguistic Knowledge Acquisition and Representation. [find](https://scholar.google.com/scholar?q=A%2E+Lopez%2C+M%2E+Nosal%2C+R%2E+Hwa%2C+and+P%2E+Resnik%2E+2002%2E+Word%2Dlevel+alignment+for+multilingual+resource+acquisition%2E+In+Proceedings+of+the+LREC%2D02+Workshop+on+Linguistic+Knowledge+Acquisition+and+Representati)
5. I. D. Melamed. 1998. Empirical methods for MT lexicon development. Lecture Notes in Computer Science, 1529:18–9999. [find](https://scholar.google.com/scholar?q=I%2E+D%2E+Melamed%2E+1998%2E+Empirical+methods+for+MT+lexicon+development%2E+Lecture+Notes+in+Computer+Science%2C+1529%3A18%E2%80%939999%2E)
6. R. Mihalcea and T. Pedersen. 2003. [An evaluation exercise for word alignment](https://md.research.elamur.ai/free/en/papers/mihalcea-2003-evaluation-exercise-word-alignment/). In Rada Mihalcea and Ted Pedersen, editors, Proceedings of the HLT-NAACL 2003 Workshop on Building and Using Parallel Texts, pages 1–10.
7. F. J. Och and H. Ney. 2000. [A comparison of alignment models for statistical machine translation](https://md.research.elamur.ai/free/en/papers/och-2000-comparison-alignment-models-statistical-machine/). In Proceedings of COLING-00, pages 1086–1090.
8. D. I. Tufiş. 2002. A cheap and fast way to build useful translation lexicons. In Proceedings of COLING-02, pages 1030–1036. [find](https://scholar.google.com/scholar?q=D%2E+I%2E+Tufis%CC%A7%2E+2002%2E+A+cheap+and+fast+way+to+build+useful+translation+lexicons%2E+In+Proceedings+of+COLING%2D02%2C+pages+1030%E2%80%931036%2E)
9. D. Wu. 1995. Stochastic inversion transduction grammars, with application to segmentation, bracketing, and alignment of parallel corpora. In Proceedings of IJCAI-95, pages 1328–1335. [find](https://scholar.google.com/scholar?q=D%2E+Wu%2E+1995%2E+Stochastic+inversion+transduction+grammars%2C+with+application+to+segmentation%2C+bracketing%2C+and+alignment+of+parallel+corpora%2E+In+Proceedings+of+IJCAI%2D95%2C+pages+1328%E2%80%931335%2E)
10. K. Yamada and K. 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 ACL-01, pages 523–530.
11. D. Yarowsky, G. Ngai, and R. 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 HLT-01, pages 161–168.

*Updated 31 October 2026*