---
title: "A Feedback-Augmented Method for Detecting Errors in the"
description: "The paper targets a frequent weakness of Japanese learners of English: errors with articles and with… 18 references, cited by 22 papers here."
authors: ["Ryo Nagata", "Atsuo Kawai", "Koichiro Morihiro", "Naoki Isu"]
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/P06-1031/"
doi: "10.3115/1220175.1220206"
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/nagata-2006-feedback-augmented-method-detecting-errors/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/nagata-2006-feedback-augmented-method-detecting-errors/)

[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/) / A Feedback-Augmented Method for Detecting Errors in the Writing of Learners of English

# A Feedback-Augmented Method for Detecting Errors in the Writing of Learners of English

**Authors:** [Ryo Nagata](https://md.research.elamur.ai/free/en/authors/ryo-nagata/), [Atsuo Kawai](https://md.research.elamur.ai/free/en/authors/atsuo-kawai/), [Koichiro Morihiro](https://md.research.elamur.ai/free/en/authors/koichiro-morihiro/) and [Naoki Isu](https://md.research.elamur.ai/free/en/authors/naoki-isu/) · **Year:** 2006 · **Venue:** [Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics](https://md.research.elamur.ai/free/en/venues/acl/) ([ACL 2006](https://md.research.elamur.ai/free/en/venues/acl/2006/)) · **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/P06-1031/) · [PDF](https://aclanthology.org/P06-1031.pdf) · [DOI 10.3115/1220175.1220206](https://doi.org/10.3115/1220175.1220206) · [OpenAlex](https://openalex.org/W2148474294)

**Citations:** 48 citations (OpenAlex, 2026-10-14) · [22](https://md.research.elamur.ai/free/en/papers/nagata-2006-feedback-augmented-method-detecting-errors/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
The paper targets a frequent weakness of Japanese learners of English: errors with articles and with singular/plural forms. These are hard to catch with hand-written rules because whether a noun behaves as mass or count depends on its sense and context. The authors classify each noun as mass or count and then apply grammatical error rules to that label. Training examples are labelled automatically from a general corpus (BNC or EDR) using grammar-based cues such as plural form or determiners like *much* and *another*. A decision list is then learned per noun from the surrounding words. Because general text differs from learner writing, the system also uses a "feedback corpus" of learner essays corrected by a teacher. This small corpus is combined with the large one so that the large corpus does not swamp it. With feedback, the method reaches recall 0.71 and precision 0.72 and beats rule-based and web-based baselines. The work shows that cheap automatic labelling plus a small amount of in-domain correction can handle a context-dependent lexical distinction.

## Key points
- Training labels come from hand-made tagging rules grounded in Huddleston and Pullum (2002), such as treating a plural as count. About 30% of instances get no label ("?") and are dropped.
- Each noun has its own decision list. Rule conditions are words in the noun phrase, to its left, or to its right, after stemming and lowercasing, with function words removed except prepositions.
  - Rules are ordered by Yarowsky-style log-likelihood ratio, smoothed with α = 1.
  - A default majority-class rule ends the list, and anything ranked below it is discarded.
- Errors are detected in three ordered steps:
  - first, a mass noun in plural form is flagged;
  - second, determiner and number combinations from Table 3 are checked;
  - third, the further rules in Table 4 are applied.
- Results:
  - The best configuration, DL (EDR) with feedback, scores recall 0.71 and precision 0.72.
  - The feedback corpus alone gives 0.43 and 0.76, the rule-based baseline 0.59 and 0.39, and the web-based baseline 0.49 and 0.53.
  - Simply appending the feedback corpus to the general one (EDR+FB, BNC+FB) helped little or not at all.
- The best system still made 30 false negatives and 29 false positives. Roughly 70% of the misses would need information beyond the mass/count label, for example a superfluous *the*.

## Interesting ideas
- Unambiguous surface cues (plural morphology, determiners like *much*, *a*, *another*) auto-label the easy occurrences of a word, and a context model trained on them then resolves the ambiguous rest.
- Per-word decision lists ordered by log-likelihood ratio keep every decision traceable to a single context word, and rules ranked below the always-firing default rule are pruned as dead.
- A small in-domain corpus is combined with a large general corpus through log-damped counts rather than plain concatenation, so the small corpus still has influence.
- The error analysis checks which misses could never be fixed even with a perfect mass/count label, which sets a ceiling on recall gains from better classification.
- As future work, the authors propose diffing original essays against their teacher corrections to harvest error instances and to raise or lower each rule's weight based on its track record.

## For Mova

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

- **I1** (world): The same bootstrapping could give the measures and quantities layer a context-sensitive mass/count flag per noun occurrence, so the world model can tell an amount of a substance from a number of discrete objects.
- **I2** (tokens): A lexical-feature tagger in crate en could use such lists to output a decision together with the exact rule that fired, which fits the requirement that every step be explainable.
- **I3** (learn): When mixing a little teacher-corrected or LLM-tier data with large silver treebanks, damping the general counts could keep the high-quality data from being drowned out.
- **I4** (learn): Evaluations could tag each failure by whether the module under test could have prevented it, so effort goes to stages whose upstream input is the real bottleneck.
- **I5** (learn): Self-induced repair rules could be scored against before/after correction pairs, and weak rules demoted automatically based on observed hits and false alarms.

*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

- [Hyogo University of Teacher Education](https://md.research.elamur.ai/free/en/universities/hyogo-university-of-teacher-education/) — University, Katō, Japan · 4 works · 117 citations
- [Mie University](https://md.research.elamur.ai/free/en/universities/mie-university/) — University, Tsu, Japan · 2 works · 77 citations

📚 Shelves (6): [🖥️ 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 · [💭 Psychology](https://md.research.elamur.ai/free/en/shelves/psychology/) · 1,552 works · [💭 Developmental and Educational Psychology](https://md.research.elamur.ai/free/en/shelves/psychology/developmental-and-educational-psychology/) · 499 works

🧭 Topics (1): [English](https://md.research.elamur.ai/free/en/topics/english/) · 2,263 works

## Cite

Cite as: Ryo Nagata, Atsuo Kawai, Koichiro Morihiro and Naoki Isu. 2006. A Feedback-Augmented Method for Detecting Errors in the Writing of Learners of English. Proceedings of the 21st International Conference on Computational Linguistics and 44th 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{nagata2006feedback,
  title = {{A Feedback-Augmented Method for Detecting Errors in the Writing of Learners of English}},
  author = {Ryo Nagata, Atsuo Kawai, Koichiro Morihiro and Naoki Isu},
  booktitle = {Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics},
  year = {2006},
  doi = {10.3115/1220175.1220206},
  url = {https://doi.org/10.3115/1220175.1220206}
}
```

- [← previous The Notion of Argument in Prepositional Phrase Attachment](https://md.research.elamur.ai/free/en/papers/merlo-2006-notion-argument-prepositional-phrase-attachment/)
- [next → Integrated Morphological and Syntactic Disambiguation for Modern Hebrew](https://md.research.elamur.ai/free/en/papers/tsarfaty-2006-integrated-morphological-syntactic-disambiguation-modern/)
- [↑ back to papers of 2006](https://md.research.elamur.ai/free/en/papers/2006/)

## References

18 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. K. Allan. 1980. Nouns and countability. J. Linguistic Society of America, 56(3):541–567. [↗](https://doi.org/10.2307/414449)
2. F. Bond. 2005. Translating the Untranslatable. CSLI publications, Stanford. [find](https://scholar.google.com/scholar?q=F%2E+Bond%2E+2005%2E+Translating+the+Untranslatable%2E+CSLI+publications%2C+Stanford%2E)
3. L. Burnard. 1995. Users Reference Guide for the British National Corpus. version 1.0. Oxford University Computing Services, Oxford. [find](https://scholar.google.com/scholar?q=L%2E+Burnard%2E+1995%2E+Users+Reference+Guide+for+the+British+National+Corpus%2E+version+1%2E0%2E+Oxford+University+Computing+Services%2C+Oxford%2E)
4. M. Chodorow and C. Leacock. 2000. An unsupervised method for detecting grammatical errors. In Proc. of 1st Meeting of the North America Chapter of ACL, pages 140–147. [find](https://scholar.google.com/scholar?q=M%2E+Chodorow+and+C%2E+Leacock%2E+2000%2E+An+unsupervised+method+for+detecting+grammatical+errors%2E+In+Proc%2E+of+1st+Meeting+of+the+North+America+Chapter+of+ACL%2C+pages+140%E2%80%93147%2E)
5. Japan electronic dictionary research institute ltd. 1993. EDR electronic dictionary specifications guide. Japan electronic dictionary research institute ltd, Tokyo. [find](https://scholar.google.com/scholar?q=Japan+electronic+dictionary+research+institute+ltd%2E+1993%2E+EDR+electronic+dictionary+specifications+guide%2E+Japan+electronic+dictionary+research+institute+ltd%2C+Tokyo%2E)
6. S. Granger. 1998. Prefabricated patterns in advanced EFL writing: collocations and formulae. In A. P. Cowie, editor, Phraseology: theory, analysis, and applications, pages 145–160. Clarendon Press. [↗](https://doi.org/10.1093/oso/9780198294252.003.007)
7. R. Huddleston and G.K. Pullum. 2002. [The Cambridge Grammar of the English Language](https://md.research.elamur.ai/free/en/papers/huddleston-2002-cambridge-grammar-english-language/). Cambridge University Press, Cambridge. [↗](https://doi.org/10.1017/9781316423530)
8. E. Izumi, K. Uchimoto, T. Saiga, T. Supnithi, and H. Isahara. 2003. Automatic error detection in the Japanese learners’ English spoken data. In Proc. of 41st Annual Meeting of ACL, pages 145–148. [find](https://scholar.google.com/scholar?q=E%2E+Izumi%2C+K%2E+Uchimoto%2C+T%2E+Saiga%2C+T%2E+Supnithi%2C+and+H%2E+Isahara%2E+2003%2E+Automatic+error+detection+in+the+Japanese+learners%E2%80%99+English+spoken+data%2E+In+Proc%2E+of+41st+Annual+Meeting+of+ACL%2C+pages+145%E2%80%93148%2E)
9. A. Kawai, K. Sugihara, and N. Sugie. 1984. ASPEC-I: An error detection system for English composition. IPSJ Journal (in Japanese), 25(6):1072–1079. [find](https://scholar.google.com/scholar?q=A%2E+Kawai%2C+K%2E+Sugihara%2C+and+N%2E+Sugie%2E+1984%2E+ASPEC%2DI%3A+An+error+detection+system+for+English+composition%2E+IPSJ+Journal+%28in+Japanese%29%2C+25%286%29%3A1072%E2%80%931079%2E)
10. M. Lapata and F. Keller. 2005. Web-based models for natural language processing. ACM Transactions on Speech and Language Processing, 2(1):1–31. [↗](https://doi.org/10.1145/1075389.1075392)
11. J. Lee. 2004. Automatic article restoration. In Proc. of the Human Language Technology Conference of the North American Chapter of ACL, pages 31–36. [find](https://scholar.google.com/scholar?q=J%2E+Lee%2E+2004%2E+Automatic+article+restoration%2E+In+Proc%2E+of+the+Human+Language+Technology+Conference+of+the+North+American+Chapter+of+ACL%2C+pages+31%E2%80%9336%2E)
12. K.F. McCoy, C.A. Pennington, and L.Z. Suri. 1996. English error correction: A syntactic user model based on principled “mal-rule” scoring. In Proc. of 5th International Conference on User Modeling, pages 69–66. [find](https://scholar.google.com/scholar?q=K%2EF%2E+McCoy%2C+C%2EA%2E+Pennington%2C+and+L%2EZ%2E+Suri%2E+1996%2E+English+error+correction%3A+A+syntactic+user+model+based+on+principled+%E2%80%9Cmal%2Drule%E2%80%9D+scoring%2E+In+Proc%2E+of+5th+International+Conference+on+User+Modeling%2C+pa)
13. G. Minnen, F. Bond, and A. Copestake. 2000. Memory-based learning for article generation. In Proc. of CoNLL-2000 and LLL-2000 workshop, pages 43–48. [find](https://scholar.google.com/scholar?q=G%2E+Minnen%2C+F%2E+Bond%2C+and+A%2E+Copestake%2E+2000%2E+Memory%2Dbased+learning+for+article+generation%2E+In+Proc%2E+of+CoNLL%2D2000+and+LLL%2D2000+workshop%2C+pages+43%E2%80%9348%2E)
14. N. Ostler and B.T.S Atkins. 1991. Predictable meaning shift: Some linguistic properties of lexical implication rules. In Proc. of 1st SIGLEX Workshop on Lexical Semantics and Knowledge Representation, pages 87–100. [find](https://scholar.google.com/scholar?q=N%2E+Ostler+and+B%2ET%2ES+Atkins%2E+1991%2E+Predictable+meaning+shift%3A+Some+linguistic+properties+of+lexical+implication+rules%2E+In+Proc%2E+of+1st+SIGLEX+Workshop+on+Lexical+Semantics+and+Knowledge+Representation%2C)
15. D. Schneider and K.F. McCoy. 1998. Recognizing syntactic errors in the writing of second language learners. In Proc. of 17th International Conference on Computational Linguistics, pages 1198–1205. [find](https://scholar.google.com/scholar?q=D%2E+Schneider+and+K%2EF%2E+McCoy%2E+1998%2E+Recognizing+syntactic+errors+in+the+writing+of+second+language+learners%2E+In+Proc%2E+of+17th+International+Conference+on+Computational+Linguistics%2C+pages+1198%E2%80%931205%2E)
16. Y. Tanaka. 1977. Psychological methods (in Japanese). University of Tokyo Press. œ  C. Tschichold, F. Bodmer, E. Cornu, F. Grosjean, L. Grosjean, N. K ubler, N. L ewy, and C. Tschumi. 1997. Developing a new grammar checker for English as a second language. In Proc. of the From Research to Commercial Applications Workshop, pages 7–12. [find](https://scholar.google.com/scholar?q=Y%2E+Tanaka%2E+1977%2E+Psychological+methods+%28in+Japanese%29%2E+University+of+Tokyo+Press%2E+%C2%9C+%C2%9D+C%2E+Tschichold%2C+F%2E+Bodmer%2C+E%2E+Cornu%2C+F%2E+Grosjean%2C+L%2E+Grosjean%2C+N%2E+K+ubler%2C+N%2E+L+ewy%2C+and+C%2E+Tschumi%2E+1997%2E+Developin)
17. D. Yarowsky. 1995. [Unsupervised word sense disambiguation rivaling supervised methods](https://md.research.elamur.ai/free/en/papers/yarowsky-1995-unsupervised-word-sense-disambiguation-rivaling/). In Proc. of 33rd Annual Meeting of ACL, pages 189–196.
18. D. Yarowsky. 1996. Homograph Disambiguation in Speech Synthesis. Springer-Verlag. [find](https://scholar.google.com/scholar?q=D%2E+Yarowsky%2E+1996%2E+Homograph+Disambiguation+in+Speech+Synthesis%2E+Springer%2DVerlag%2E)

*Updated 1 November 2026*