---
title: "Incorporating an Error Corpus into a Spellchecker for Maltese — Rosner 2012"
description: "The paper addresses the lack of a dependable spellchecker for Maltese. 0.620 recall. 16 references, cited by 1 paper here."
authors: ["Michael Rosner", "Albert Gatt", "Andrew Attard", "Jan Joachimsen"]
year: 2012
venue: "Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC'12)"
source: "https://aclanthology.org/L12-1620/"
licence: "CC BY-NC 4.0 (ELRA, LREC proceedings)"
licence_url: "https://creativecommons.org/licenses/by-nc/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/rosner-2012-incorporating-error-corpus-into-spellchecker/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/rosner-2012-incorporating-error-corpus-into-spellchecker/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2012](https://md.research.elamur.ai/free/en/papers/2012/page/2/) / Incorporating an Error Corpus into a Spellchecker for Maltese

# Incorporating an Error Corpus into a Spellchecker for Maltese

**Authors:** [Michael Rosner](https://md.research.elamur.ai/free/en/authors/michael-rosner/), [Albert Gatt](https://md.research.elamur.ai/free/en/authors/albert-gatt/), Andrew Attard and Jan Joachimsen · **Year:** 2012 · **Venue:** Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC'12) · **Licence:** [CC BY-NC 4.0 (ELRA, LREC proceedings)](https://creativecommons.org/licenses/by-nc/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/L12-1620/) · [PDF](https://aclanthology.org/L12-1620.pdf) · [OpenAlex](https://openalex.org/W2250305284)

**Citations:** 0 citations (OpenAlex, 2026-10-14) · 1 in our library, of the 140,884 papers whose bibliographies we read

## Summary
The paper addresses the lack of a dependable spellchecker for Maltese. The language is hard for this task because it mixes Semitic root-and-pattern morphology with stem-based loanword morphology, has many spelling variants, and has few resources. The authors review earlier systems: a tiny-lexicon edit-operation approach, a lexicon-free n-gram filter, and an aspell-based checker with an 800K-type wordlist. They argue these systems ignore context and that their quality was never measured. Two new resources make a better approach possible: the c. 100M-word MLRS corpus and a detailed corpus of spelling errors collected while MLRS texts were being cleaned. The authors outline a pluggable framework with consistent evaluation criteria. They also build a baseline that checks isolated words against a wordlist and proposes candidates one edit away, ranked by MLRS frequency. Over ten randomized test runs the baseline reaches recall around 0.93–0.97 but precision only about 0.58–0.62 (F ≈ 0.72–0.75). This gives Maltese its first measured benchmark for future context-aware and morphology-aware improvements.

## Key points
- Resources: the Maltese Language Resource Server (MLRS) corpus (~100M words from legislation, press releases and web pages) and a hand-corrected error corpus gathered during MLRS text preparation.
- The error corpus shows three error sources:
  - legacy text lacking Unicode Maltese letters (e.g. haga for ħaġa);
  - inconsistent spelling of loanwords;
  - ordinary typos.
- Error counts by minimum edit distance (MED) are 268,141 (MED 0), 164,665 (1), 68,962 (2), 26,626 (3), 10,876 (4) and 6,071 (5).
- The baseline system works as follows:
  - it uses a hashed lookup in R. Casha's Linux User Group wordlist for detection;
  - it takes MED = 1 candidates for correction;
  - it ranks those candidates by frequency in the MLRS corpus.
- Evaluation covers only Reynaert (2008) level 1. Each test file holds 500 words (250 correct, 250 misspelled at MED 1), repeated over 10 random draws. The best F-score was 0.752 (precision 0.620, recall 0.956).

## Interesting ideas
- Collect a spelling-error corpus as a free side effect of manually cleaning corpus text, so every human correction becomes a labelled error/fix pair.
- Build test sets that are half genuinely correct items and half known errors, and redraw them randomly over many cycles to see how stable precision and recall are.
- Stratify errors by edit distance to see which difficulty levels a corrector can and cannot reach before claiming coverage.
- Check words for well-formedness by parsing them into an abstract template (root slots plus vowels and affixes) and matching that template against the paradigm of the word's class, rather than only looking them up in a list.
- Treat encoding-degraded text (missing diacritics or special letters) as its own error class, separate from real misspellings and variant spellings.

## For Mova

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

- **I1** (learn): Log every fix made by Mova's normalizer or by human reviewers on silver data as error→fix pairs, then use them to induce and test self-repair rules.
- **I2** (learn): Use correct inputs as built-in negative controls for any repair component, and report variance across resampled test sets rather than a single score.
- **I3** (learn): Bucket Mova's tokenizer and lemma errors by how far they are from the gold form, so evaluation reveals whether failures cluster in "hard" distances.
- **I4** (tokens): Validate unseen inflected forms against class-specific paradigm templates in crate en, so the morphology module can explain why a form is accepted or rejected.
- **I5** (tokens): Add a dedicated normalization pass for encoding artefacts (smart quotes, stripped accents in names) before tagging, so these are not confused with genuine out-of-vocabulary words.

*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: [Linguistics & languages](https://md.research.elamur.ai/free/en/shelves/linguistics/)

Topics: [🧩 Morphology](https://md.research.elamur.ai/free/en/topics/morphology/) · [🐪 Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/)

## Institutions

- [University of Malta](https://md.research.elamur.ai/free/en/universities/university-of-malta/) — University, Msida, Malta · 12 works · 573 citations

## Cite

Cite as: Michael Rosner, Albert Gatt, Andrew Attard and Jan Joachimsen. 2012. Incorporating an Error Corpus into a Spellchecker for Maltese. Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC'12).

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

```bibtex
@inproceedings{rosner2012incorporating,
  title = {{Incorporating an Error Corpus into a Spellchecker for Maltese}},
  author = {Michael Rosner, Albert Gatt, Andrew Attard and Jan Joachimsen},
  booktitle = {Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC'12)},
  year = {2012},
  doi = {10.63317/22eqsp83hfhx},
  url = {https://doi.org/10.63317/22eqsp83hfhx}
}
```

- [← previous Conversion of Procedural Morphologies to Finite-State Morphologies: A Case Study of Arabic](https://md.research.elamur.ai/free/en/papers/hulden-2012-conversion-procedural-morphologies-finite-state/)
- [next → Using Arabic Transliteration to Improve Word Alignment from French- Arabic Parallel Corpora](https://md.research.elamur.ai/free/en/papers/saadane-2012-arabic-transliteration-improve-word-alignment/)
- [↑ back to papers of 2012](https://md.research.elamur.ai/free/en/papers/2012/page/2/)

## References

16 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Akkademja tal-Malti. 2001. Regoli tal-Kitba tal-Malti. Klabb Kotba Maltin, Malta: Valletta. [find](https://scholar.google.com/scholar?q=Akkademja+tal%2DMalti%2E+2001%2E+Regoli+tal%2DKitba+tal%2DMalti%2E+Klabb+Kotba+Maltin%2C+Malta%3A+Valletta%2E)
2. E. Atwell and S. Elliott. 1987. Dealing with ill-formed english text. In R. Garside, G. Leach, and G Sampson, editors, Computational Analysis of English: A Corpus-Based Approach, chapter 10. Longmans, New York. [find](https://scholar.google.com/scholar?q=E%2E+Atwell+and+S%2E+Elliott%2E+1987%2E+Dealing+with+ill%2Dformed+english+text%2E+In+R%2E+Garside%2C+G%2E+Leach%2C+and+G+Sampson%2C+editors%2C+Computational+Analysis+of+English%3A+A+Corpus%2DBased+Approach%2C+chapter+10%2E+Longmans%2C)
3. F. Damerau and E. Mays. 1989. An examination of undetected typing errors. Inf. Process. Manage., 25(6):659– 664. [find](https://scholar.google.com/scholar?q=F%2E+Damerau+and+E%2E+Mays%2E+1989%2E+An+examination+of+undetected+typing+errors%2E+Inf%2E+Process%2E+Manage%2E%2C+25%286%29%3A659%E2%80%93+664%2E)
4. R. Fabri. 2009. Stem allomorphy in the Maltese verb. In T. Stolz, editor, Ilsienna - Our Language, volume 1, pages 1–20. Dr. N Brokmeyer, Bochum. [find](https://scholar.google.com/scholar?q=R%2E+Fabri%2E+2009%2E+Stem+allomorphy+in+the+Maltese+verb%2E+In+T%2E+Stolz%2C+editor%2C+Ilsienna+%2D+Our+Language%2C+volume+1%2C+pages+1%E2%80%9320%2E+Dr%2E+N+Brokmeyer%2C+Bochum%2E)
5. Il-Kunsill Nazzjonali tal-Ilsien Malti. 2008. Deċiżjonijiet 1 tal-kunsill nazzjonali tal-ilsien malti dwar il-varjanti ortografiċi. Il-Furjana Press. [find](https://scholar.google.com/scholar?q=Il%2DKunsill+Nazzjonali+tal%2DIlsien+Malti%2E+2008%2E+Dec%CC%87iz%CC%87jonijiet+1+tal%2Dkunsill+nazzjonali+tal%2Dilsien+malti+dwar+il%2Dvarjanti+ortografic%CC%87i%2E+Il%2DFurjana+Press%2E)
6. Mark D. Kernighan, Kenneth Church, and William A. Gale. 1990. A spelling correction program based on a noisy channel model. In Proceedings of the Thirteenth International Conference on Computational Linguistics, pages 205–210. http://acl.ldc.upenn.edu/C/C90/C90-2036.pdf. [↗](https://aclanthology.org/C90-2036/)
7. Adam Kilgarriff, Pavel Rychly, Pavel Smrz, and David Tugwell. 2004. The sketch engine. In Proceedings of EURALEX. [find](https://scholar.google.com/scholar?q=Adam+Kilgarriff%2C+Pavel+Rychly%2C+Pavel+Smrz%2C+and+David+Tugwell%2E+2004%2E+The+sketch+engine%2E+In+Proceedings+of+EURALEX%2E)
8. Karen Kukich. 1992. Techniques for automatically correcting words in text. ACM Comput. Surv., 24:377–439, December. http://doi.acm.org/10.1145/146370.146380. [↗](https://doi.org/10.1145/146370.146380)
9. G. Mangion. 1999. Spelling Correction for Maltese. Technical report, Dept. CSAI, University of Malta, Msida MSD2080, Malta. [find](https://scholar.google.com/scholar?q=G%2E+Mangion%2E+1999%2E+Spelling+Correction+for+Maltese%2E+Technical+report%2C+Dept%2E+CSAI%2C+University+of+Malta%2C+Msida+MSD2080%2C+Malta%2E)
10. M. McIlroy. 1982. Development of a spelling list. IEEE Transactions Communcations, 30(1):91 – 99, jan. [find](https://scholar.google.com/scholar?q=M%2E+McIlroy%2E+1982%2E+Development+of+a+spelling+list%2E+IEEE+Transactions+Communcations%2C+30%281%29%3A91+%E2%80%93+99%2C+jan%2E)
11. R. Mitton. 1987. Spelling Checkers, Spelling Correctors, and the Misspellings of Poor Spellers. Inf. Process. Manag., 23(5):495–505. [find](https://scholar.google.com/scholar?q=R%2E+Mitton%2E+1987%2E+Spelling+Checkers%2C+Spelling+Correctors%2C+and+the+Misspellings+of+Poor+Spellers%2E+Inf%2E+Process%2E+Manag%2E%2C+23%285%29%3A495%E2%80%93505%2E)
12. R. Mizzi. 2000. The Development of a Statistical Spell Checker for Maltese. Technical report, Dept. CSAI, University of Malta, Msida MSD2080, Malta. [find](https://scholar.google.com/scholar?q=R%2E+Mizzi%2E+2000%2E+The+Development+of+a+Statistical+Spell+Checker+for+Maltese%2E+Technical+report%2C+Dept%2E+CSAI%2C+University+of+Malta%2C+Msida+MSD2080%2C+Malta%2E)
13. J. Peterson. 1986. A Note on Undetected Typing Errors. CACM, 29(7):633–637. [find](https://scholar.google.com/scholar?q=J%2E+Peterson%2E+1986%2E+A+Note+on+Undetected+Typing+Errors%2E+CACM%2C+29%287%29%3A633%E2%80%93637%2E)
14. M Reynaert. 2008. All, and only, the errors: more complete and consistent spelling and ocr-error correction evaluation. In N. Calzolari, K. Choukri, B. Maegaard, J. Mariani, J. Odjik, S. Piperidis, and D. Tapias, editors, Proc. LREC 2008, Marrakech, Morocco. European Language Resources Association (ELRA). [http://ilk.uvt.nl/downloads/pub/papers/477_paper.pdf](http://ilk.uvt.nl/downloads/pub/papers/477_paper.pdf). [↗](http://ilk.uvt.nl/downloads/pub/papers/477_paper.pdf)
15. G. 1989 Sampson. 1989. How fully does a machine-usable dictionary cover english text. Lit. Ling Computing, 4(1):29–35. [find](https://scholar.google.com/scholar?q=G%2E+1989+Sampson%2E+1989%2E+How+fully+does+a+machine%2Dusable+dictionary+cover+english+text%2E+Lit%2E+Ling+Computing%2C+4%281%29%3A29%E2%80%9335%2E)
16. A. Wing and A. Baddeley. 1980. Spelling errors in handwriting: A corpus and distributional analysis. In U. Frith, editor, Cognitive Processes in Spelling. Academic Press, London. [find](https://scholar.google.com/scholar?q=A%2E+Wing+and+A%2E+Baddeley%2E+1980%2E+Spelling+errors+in+handwriting%3A+A+corpus+and+distributional+analysis%2E+In+U%2E+Frith%2C+editor%2C+Cognitive+Processes+in+Spelling%2E+Academic+Press%2C+London%2E)