---
title: "Correcting ESL Errors Using Phrasal SMT — Brockett 2006"
description: "The paper reframes correction of English-learner (ESL) writing as machine translation from flawed learner English into native-like English."
authors: ["Chris Brockett", "William B. Dolan", "Michael Gamon"]
year: 2006
venue: "Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics"
source: "https://aclanthology.org/P06-1032/"
doi: "10.3115/1220175.1220207"
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/brockett-2006-correcting-esl-errors-phrasal-smt/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/brockett-2006-correcting-esl-errors-phrasal-smt/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2006](https://md.research.elamur.ai/free/en/papers/2006/) / Correcting ESL Errors Using Phrasal SMT Techniques

# Correcting ESL Errors Using Phrasal SMT Techniques

**Authors:** [Chris Brockett](https://md.research.elamur.ai/free/en/authors/chris-brockett/), [William B. Dolan](https://md.research.elamur.ai/free/en/authors/william-b-dolan/), [Michael Gamon](https://md.research.elamur.ai/free/en/authors/michael-gamon/) · **Year:** 2006 · **Venue:** Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the 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/P06-1032/) · [PDF](https://aclanthology.org/P06-1032.pdf) · [DOI 10.3115/1220175.1220207](https://doi.org/10.3115/1220175.1220207) · [OpenAlex](https://openalex.org/W2123388068)

**Citations:** 220 citations (OpenAlex, 2026-10-14) · [99](https://md.research.elamur.ai/free/en/papers/brockett-2006-correcting-esl-errors-phrasal-smt/cited-by/) in our library, of the 140884 papers whose bibliographies we read

## Summary
The paper reframes correction of English-learner (ESL) writing as machine translation from flawed learner English into native-like English. This lets a system propose fixes for whole phrases instead of only flagging problems. As a pilot, the authors target countability errors with mass nouns, which are common among writers with East Asian L1s. They select 14 such nouns using dictionaries and the Chinese Learner English Corpus (CLEC). Since no large parallel corpus of learner and corrected text existed, they built one artificially. They took well-formed newswire sentences and used regular expressions to inject CLEC-style errors, then mixed these with unchanged sentences. They trained Microsoft's dependency-treelet phrasal SMT system on this data without changing its architecture. On real mass-noun errors collected from the Web, the system fixed about 62% of mistakes, far outperforming Word 2003's built-in grammar and spelling checker. The result suggests that collecting genuine before/after editing corpora could enable data-driven writing assistants for many kinds of learner errors.

## Key points
- Target vocabulary: 14 mass nouns, including knowledge, information, advice, homework, news, equipment and tea. Each was listed with a mass sense in Longman or American Heritage and appeared in at least 2 mass/count errors in CLEC.
- Baseline on the 46 relevant CLEC sentences: Word 2003 caught 21 errors, made one wrong suggestion and missed 25. It handled forms like *informations* but ignored plurals such as *fruits* or *papers* used in the wrong context.
- Training data started from about 484.6M words of Reuters newswire (1995–1998) plus about 7.2M words of 2004–2005 news. From this, roughly 346,000 sentences containing the target nouns were selected. Of these, 24,128 were "lesioned" into 65,826 erroneous/correct pairs and balanced with an equal share of identity pairs. Training sets went up to 45K pairs, and 1,000 lesioned sentences were held out for lambda training via Maximum Bleu.
- The system maps dependency "treelets" to target strings (Tree-To-String). This lets it model non-adjacent relations such as determiner–noun links while tolerating ill-formed input.
- With 45K training pairs, 55.28% of web test sentences were fully corrected, compared with 29.27% for Word. Word missed 57.72% of errors, versus 21.14% for the system. Smaller training sets performed worse.

## Interesting ideas
- **Synthetic error injection guided by a small learner corpus.** A handful of real, annotated learner errors can serve as a template for corrupting large amounts of clean text. This gets around the scarcity of parallel correction data and could be reused for any low-resource language or for error types that are rarely tagged.
- **Mixing in identity pairs.** Including sentences that map to themselves keeps the model from learning that a given word must always be rewritten. This simple guard against over-correction applies to any artificial-noise training setup.
- **Error correction as noisy-channel translation.** Treating learner text as native text distorted by L1 interference gives a mechanism that proposes rewrites, not just flags. It can also be retrained for different L1 backgrounds or proficiency levels simply by swapping the data.
- **Finer-grained evaluation categories.** Outputs are scored per sentence as whole fix, partial fix, correctly left unchanged, new error introduced, missed, or word-order error. This shows the precision and over-correction tradeoffs that a single accuracy number would hide.
- **Reusing hidden editing workflows as data sources.** Teachers, corporate language helpdesks and translation bureaus already produce pre- and post-editing text pairs that are usually thrown away. Collecting them systematically could supply training data for correction models.

*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](https://md.research.elamur.ai/free/en/shelves/nlp/)

Topics: [Machine translation](https://md.research.elamur.ai/free/en/topics/machine-translation/) · [Grammatical error correction & learner language](https://md.research.elamur.ai/free/en/topics/grammatical-error-correction/)

## Cite

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

```bibtex
@inproceedings{brockett2006correcting,
  title = {{Correcting ESL Errors Using Phrasal SMT Techniques}},
  author = {Brockett, Chris and Dolan, William B. and Gamon, Michael},
  booktitle = {Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics},
  year = {2006},
  doi = {10.3115/1220175.1220207},
  url = {https://doi.org/10.3115/1220175.1220207}
}
```

- [← previous Paraphrasing for Automatic Evaluation](https://md.research.elamur.ai/free/en/papers/kauchak-2006-paraphrasing-automatic-evaluation/)
- [next → N-gram-based Machine Translation](https://md.research.elamur.ai/free/en/papers/marino-2006-n-gram-based-machine-translation/)
- [↑ back to papers of 2006](https://md.research.elamur.ai/free/en/papers/2006/)

## References

19 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Bond, Francis, Kentaro Ogura and Satoru Ikehara. 1994. Countability and Number in Japanese-to-English Machine Translation. COLING-94. [find](https://scholar.google.com/scholar?q=Bond%2C+Francis%2C+Kentaro+Ogura+and+Satoru+Ikehara%2E+1994%2E+Countability+and+Number+in+Japanese%2Dto%2DEnglish+Machine+Translation%2E+COLING%2D94%2E)
2. Peter E Brown, Stephen A. Della Pietra, Robert L. Mercer, and Vincent J. Della Pietra. 1993. The Mathematics of Statistical Machine Translation. Computational Linguistics, Vol. 19(2): 263-311. [find](https://scholar.google.com/scholar?q=Peter+E+Brown%2C+Stephen+A%2E+Della+Pietra%2C+Robert+L%2E+Mercer%2C+and+Vincent+J%2E+Della+Pietra%2E+1993%2E+The+Mathematics+of+Statistical+Machine+Translation%2E+Computational+Linguistics%2C+Vol%2E+19%282%29%3A+263%2D311%2E)
3. Martin Chodorow and Claudia Leacock. 2000. An Unsupervised Method for Detecting Grammatical Errors. [find](https://scholar.google.com/scholar?q=Martin+Chodorow+and+Claudia+Leacock%2E+2000%2E+An+Unsupervised+Method+for+Detecting+Grammatical+Errors%2E)
4. Michael Collins, Philipp Koehn and Ivona Kučerová. 2005. Clause Restructuring for Statistical machine Translation. [find](https://scholar.google.com/scholar?q=Michael+Collins%2C+Philipp+Koehn+and+Ivona+Ku%C4%8Derov%C3%A1%2E+2005%2E+Clause+Restructuring+for+Statistical+machine+Translation%2E) [our card](https://md.research.elamur.ai/free/en/papers/collins-2005-clause-restructuring-statistical-machine-translation/)
5. Gerard M. Dalgish. 1984. Computer-Assisted ESL Research. CALICO Journal. 2(2): 32-33 [find](https://scholar.google.com/scholar?q=Gerard+M%2E+Dalgish%2E+1984%2E+Computer%2DAssisted+ESL+Research%2E+CALICO+Journal%2E+2%282%29%3A+32%2D33)
6. Heidi J. Fox. 2002. Phrasal Cohesion and Statistical Machine Translation. [find](https://scholar.google.com/scholar?q=Heidi+J%2E+Fox%2E+2002%2E+Phrasal+Cohesion+and+Statistical+Machine+Translation%2E)
7. Shicun Gui and Huizhong Yang (eds). 2003 Zhongguo Xuexizhe Yingyu Yuliaohu. (Chinese Learner English Corpus). Shanghai: Shanghai Waiyu Jiaoyu Chubanshe. (In Chinese). [find](https://scholar.google.com/scholar?q=Shicun+Gui+and+Huizhong+Yang+%28eds%29%2E+2003+Zhongguo+Xuexizhe+Yingyu+Yuliaohu%2E+%28Chinese+Learner+English+Corpus%29%2E+Shanghai%3A+Shanghai+Waiyu+Jiaoyu+Chubanshe%2E+%28In+Chinese%29%2E)
8. Hua Dongfan and Thomas Hun-Tak Lee. 2004. Chinese ESL Learners' Understanding of the English Count-Mass Distinction. In Proceedings of the 7th Generative Approaches to Second Language Acquisition Conference (GASLA 2004). [find](https://scholar.google.com/scholar?q=Hua+Dongfan+and+Thomas+Hun%2DTak+Lee%2E+2004%2E+Chinese+ESL+Learners%27+Understanding+of+the+English+Count%2DMass+Distinction%2E+In+Proceedings+of+the+7th+Generative+Approaches+to+Second+Language+Acquisition+Conf)
9. Ting Liu, Ming Zhou, Jianfeng Gao, Endong Xun, and Changning Huang. 2000. PENS: A Machine-aided English Writing System for Chinese Users. [find](https://scholar.google.com/scholar?q=Ting+Liu%2C+Ming+Zhou%2C+Jianfeng+Gao%2C+Endong+Xun%2C+and+Changning+Huang%2E+2000%2E+PENS%3A+A+Machine%2Daided+English+Writing+System+for+Chinese+Users%2E)
10. Deryle Lonsdale and Diane Strong-Krause. 2003. Automated Rating of ESL Essays. In Proceedings of the HLT/NAACL Workshop: Building Educational Applications Using Natural Language Processing. [find](https://scholar.google.com/scholar?q=Deryle+Lonsdale+and+Diane+Strong%2DKrause%2E+2003%2E+Automated+Rating+of+ESL+Essays%2E+In+Proceedings+of+the+HLT%2FNAACL+Workshop%3A+Building+Educational+Applications+Using+Natural+Language+Processing%2E)
11. Arul Menezes, and Chris Quirk. 2005. Microsoft Research Treelet Translation System: IWSLT Evaluation. Proceedings of the International Workshop on Spoken Language Translation. [find](https://scholar.google.com/scholar?q=Arul+Menezes%2C+and+Chris+Quirk%2E+2005%2E+Microsoft+Research+Treelet+Translation+System%3A+IWSLT+Evaluation%2E+Proceedings+of+the+International+Workshop+on+Spoken+Language+Translation%2E)
12. Franz Josef Och, 2003. Minimum error rate training in statistical machine translation. [find](https://scholar.google.com/scholar?q=Franz+Josef+Och%2C+2003%2E+Minimum+error+rate+training+in+statistical+machine+translation%2E)
13. Franz Josef Och and Hermann Ney. 2000. [find](https://scholar.google.com/scholar?q=Franz+Josef+Och+and+Hermann+Ney%2E+2000%2E)
14. Improved Statistical Alignment Models. ACL 2000. [find](https://scholar.google.com/scholar?q=Improved+Statistical+Alignment+Models%2E+ACL+2000%2E)
15. Chris Quirk, Arul Menezes, and Colin Cherry. 2005. [find](https://scholar.google.com/scholar?q=Chris+Quirk%2C+Arul+Menezes%2C+and+Colin+Cherry%2E+2005%2E)
16. Dependency Tree Translation: Syntactically Informed Phrasal SMT. ACL 2005. [find](https://scholar.google.com/scholar?q=Dependency+Tree+Translation%3A+Syntactically+Informed+Phrasal+SMT%2E+ACL+2005%2E)
17. Veit Reuer. 2003. Error Recognition and Feedback with Lexical Functional Grammar. CALICO Journal, 20(3): 497-512. [find](https://scholar.google.com/scholar?q=Veit+Reuer%2E+2003%2E+Error+Recognition+and+Feedback+with+Lexical+Functional+Grammar%2E+CALICO+Journal%2C+20%283%29%3A+497%2D512%2E)
18. Laura Mayfield Tomokiyo and Rosie Jones. 2001. You’re not from round here, are you? Naive Bayes Detection of Non-Native Utterance Text. [find](https://scholar.google.com/scholar?q=Laura+Mayfield+Tomokiyo+and+Rosie+Jones%2E+2001%2E+You%E2%80%99re+not+from+round+here%2C+are+you%3F+Naive+Bayes+Detection+of+Non%2DNative+Utterance+Text%2E)
19. Anne Vandeventer Faltin. 2003. Natural language processing tools for computer assisted language learning. Linguistik online 17, 5/03 (http://www.linguistik-online.de/17_03/vandeventer.html) [find](https://scholar.google.com/scholar?q=Anne+Vandeventer+Faltin%2E+2003%2E+Natural+language+processing+tools+for+computer+assisted+language+learning%2E+Linguistik+online+17%2C+5%2F03+%28http%3A%2F%2Fwww%2Elinguistik%2Donline%2Ede%2F17%5F03%2Fvandeventer%2Ehtml%29)