---
title: "Methods for Amharic Part-of-Speech Tagging — Gambäck 2009"
description: "Amharic is the national working language of Ethiopia and has tens of millions of speakers, yet it has… 34 references, cited by 7 papers here."
authors: ["Björn Gambäck", "Fredrik Olsson", "Atelach Alemu Argaw", "Lars Asker"]
year: 2009
venue: "Proceedings of the First Workshop on Language Technologies for African Languages"
source: "https://aclanthology.org/W09-0715/"
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/gamback-2009-methods-amharic-part-speech-tagging/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/gamback-2009-methods-amharic-part-speech-tagging/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2009](https://md.research.elamur.ai/free/en/papers/2009/) / Methods for Amharic Part-of-Speech Tagging

# Methods for Amharic Part-of-Speech Tagging

**Authors:** [Björn Gambäck](https://md.research.elamur.ai/free/en/authors/bjorn-gamback/), [Fredrik Olsson](https://md.research.elamur.ai/free/en/authors/fredrik-olsson/), [Atelach Alemu Argaw](https://md.research.elamur.ai/free/en/authors/atelach-alemu-argaw/) and [Lars Asker](https://md.research.elamur.ai/free/en/authors/lars-asker/) · **Year:** 2009 · **Venue:** [Proceedings of the First Workshop on Language Technologies for African Languages](https://md.research.elamur.ai/free/en/venues/workshop-on-language-technologies-for-african-languages/) ([Workshop on Language Technologies for African Languages 2009](https://md.research.elamur.ai/free/en/venues/workshop-on-language-technologies-for-african-languages/2009/)) · **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/W09-0715/) · [PDF](https://aclanthology.org/W09-0715.pdf) · [OpenAlex](https://openalex.org/W2160055183)

**Citations:** 31 citations (OpenAlex, 2026-10-14) · [7](https://md.research.elamur.ai/free/en/papers/gamback-2009-methods-amharic-part-speech-tagging/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
Amharic is the national working language of Ethiopia and has tens of millions of speakers, yet it has very few computational resources. The authors ask how well standard supervised part-of-speech taggers work once a reasonably sized annotated corpus exists. Their data is about 200K words of ELRC-annotated Amharic text, which was the earliest tagged corpus of moderate size for the language to be released to the public. On it they train three taggers: TnT (HMM), SVMTool (SVM) and a Maximum Entropy tagger built with MALLET. Each tagger is evaluated with 10-fold cross validation on three tagsets: the full ELRC set of 30 tags and two reduced sets of 11 and 10 tags.

All three taggers score below the figures usually reported for English, and unknown words are the main source of errors. MaxEnt gives the best numbers when it is allowed to build its own folds. On shared folds the three systems perform about equally. The paper is an early benchmark for an under-resourced Semitic language. It also shows how much the choice of folds can change cross-validation results.

## Key points
- Corpus: about 200K tagged Amharic words. Unknown words occur roughly four times as often as in the Wall Street Journal corpus, which is about six times larger.
- On its own folds, MaxEnt (MALLET) reached 90.1% on the 30-tag ELRC set, and 94.6% and 94.5% on the 11-tag and 10-tag sets.
- On predefined folds, all taggers reached 92.5–92.8% on the reduced tagsets and scored 4–7 points lower on the full tagset.
- TnT was strongest on known words and weakest on unknown words. SVMTool handled unknown words best and came out slightly ahead overall. Raising TnT's suffix length from 10 to 20 gave only a marginal gain on ELRC.
- TnT trains in seconds, while MALLET and SVMTool need hours. The baseline is the share of tokens tagged as regular nouns.

## Interesting ideas
- MaxEnt accuracy jumped when MALLET chose its own folds, and per-fold variance grew at the same time, so fold choice can distort a comparison as much as the model does.
- Accuracy is reported separately for known and unknown words, and the unknown-word rate is compared with English, which explains why the overall ranking turned out as it did.
- The MaxEnt tagger represents unknown words with character prefixes and suffixes of length 1–5 plus digit, hyphen and capitalisation flags, all computed within sentence boundaries.
- The same corpus is evaluated under a fine 30-tag set and two coarse sets of about 10 tags, which shows how much accuracy the fine distinctions cost.
- The authors note that combining taggers only helps when the members are dissimilar and make different kinds of errors.

## For Mova

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

- **I1** (learn): Compare all parsers and taggers on one fixed, shared split, and report per-fold spread next to the mean so that split artefacts are not mistaken for real improvements.
- **I2** (tokens): Break tagger and lemma scores down by in-vocabulary versus OOV tokens to find out whether a regression comes from the lexicon or from the guessing component.
- **I3** (tokens): Add a compact, inspectable affix feature set as a fallback for unseen words, so that every guessed tag can be traced back to a named suffix or prefix feature.
- **I4** (tree): Score the parser on both coarse UPOS and fine-grained tags during dialect conversion, to see which fine distinctions drive attachment errors.
- **I5** (learn): Check how much the errors of the witness parsers overlap before trusting their agreement as silver data, because correlated witnesses add confidence without adding evidence.

*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

- [Stockholm University](https://md.research.elamur.ai/free/en/universities/stockholm-university/) — University, Stockholm, Sweden · 21 works · 1,832 citations
- [Swedish Institute of Computer Science](https://md.research.elamur.ai/free/en/ngos/swedish-institute-of-computer-science/) — Nonprofit, Kista, Sweden · 6 works · 649 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): [Semitic languages](https://md.research.elamur.ai/free/en/topics/semitic-languages/) · 2,908 works · [🔤 Part-of-speech tagging](https://md.research.elamur.ai/free/en/topics/part-of-speech-tagging/) · 126 works

## Cite

Cite as: Björn Gambäck, Fredrik Olsson, Atelach Alemu Argaw and Lars Asker. 2009. Methods for Amharic Part-of-Speech Tagging. Proceedings of the First Workshop on Language Technologies for African Languages.

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

```bibtex
@inproceedings{gamback2009methods,
  title = {{Methods for Amharic Part-of-Speech Tagging}},
  author = {Björn Gambäck, Fredrik Olsson, Atelach Alemu Argaw and Lars Asker},
  booktitle = {Proceedings of the First Workshop on Language Technologies for African Languages},
  year = {2009},
  doi = {10.3115/1564508.1564527},
  url = {https://doi.org/10.3115/1564508.1564527}
}
```

- [← previous Inferring Shallow-Transfer Machine Translation Rules from Small Parallel Corpora](https://md.research.elamur.ai/free/en/papers/sanchez-martinez-2014-inferring-shallow-transfer-machine-translation/)
- [next → A Unification based Approach to the Morphological Analysis and Generation of Arabic](https://md.research.elamur.ai/free/en/papers/kopru-2009-unification-based-approach-morphological-analysis/)
- [↑ back to papers of 2009](https://md.research.elamur.ai/free/en/papers/2009/)

## References

34 references, 12 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Rachel V. Xavier Aires, Sandra M. Aluı́sio, Denise C. S. Kuhn, Marcio L. B. Andreeta, and Osvaldo N. Oliveira Jr. 2000. Combining classifiers to improve part of speech tagging: A case study for Brazilian Portuguese. In 15th Brazilian Symposium on AI, pp. 227–236, Atibaia, Brazil. [find](https://scholar.google.com/scholar?q=Rachel+V%2E+Xavier+Aires%2C+Sandra+M%2E+Alu%C4%B1%CC%81sio%2C+Denise+C%2E+S%2E+Kuhn%2C+Marcio+L%2E+B%2E+Andreeta%2C+and+Osvaldo+N%2E+Oliveira+Jr%2E+2000%2E+Combining+classifiers+to+improve+part+of+speech+tagging%3A+A+case+study+for+Brazil)
2. Nega Alemayehu and Peter Willett. 2002. Stemming of Amharic words for information retrieval. Literary and Linguistic Computing, 17:1–17. [↗](https://doi.org/10.1093/llc/17.1.1)
3. Saba Amsalu and Dafydd Gibbon. 2005. Finite state morphology of Amharic. In 5th Recent Advances in Natural Language Processing, pp. 47–51, Borovets, Bulgaria. [find](https://scholar.google.com/scholar?q=Saba+Amsalu+and+Dafydd+Gibbon%2E+2005%2E+Finite+state+morphology+of+Amharic%2E+In+5th+Recent+Advances+in+Natural+Language+Processing%2C+pp%2E+47%E2%80%9351%2C+Borovets%2C+Bulgaria%2E)
4. Saba Amsalu and Girma A. Demeke. 2006. Non-concatinative finite-state morphotactics of Amharic simple verbs. ELRC Working Papers, 2:304-325. [find](https://scholar.google.com/scholar?q=Saba+Amsalu+and+Girma+A%2E+Demeke%2E+2006%2E+Non%2Dconcatinative+finite%2Dstate+morphotactics+of+Amharic+simple+verbs%2E+ELRC+Working+Papers%2C+2%3A304%2D325%2E)
5. Atelach Alemu Argaw and Lars Asker. 2007. [An Amharic stemmer: Reducing words to their citation forms](https://md.research.elamur.ai/free/en/papers/argaw-2007-amharic-stemmer-reducing-words-their/). Computational Approaches to Semitic Languages, pp. 104–110, Prague, Czech Rep.
6. Michele Banko and Robert C. Moore. 2004. Part of speech tagging in context. In 20th Int. Conf. on Computational Linguistics, pp. 556–561, Geneva, Switzerland. [find](https://scholar.google.com/scholar?q=Michele+Banko+and+Robert+C%2E+Moore%2E+2004%2E+Part+of+speech+tagging+in+context%2E+In+20th+Int%2E+Conf%2E+on+Computational+Linguistics%2C+pp%2E+556%E2%80%93561%2C+Geneva%2C+Switzerland%2E)
7. Roy Bar-Haim, Khalil Simaan, and Yoad Winter. 2008. Part-of-speech tagging of modern Hebrew text. Natural Language Engineering, 14:223–251. [↗](https://doi.org/10.1017/s135132490700455x)
8. Abiyot Bayou. 2000. Design and development of word parser for Amharic language. MSc Thesis, Addis Ababa University, Ethiopia. [find](https://scholar.google.com/scholar?q=Abiyot+Bayou%2E+2000%2E+Design+and+development+of+word+parser+for+Amharic+language%2E+MSc+Thesis%2C+Addis+Ababa+University%2C+Ethiopia%2E)
9. Tesfaye Bayu. 2002. Automatic morphological analyser: An experiment using unsupervised and autosegmental approach. MSc Thesis, Addis Ababa University, Ethiopia. [find](https://scholar.google.com/scholar?q=Tesfaye+Bayu%2E+2002%2E+Automatic+morphological+analyser%3A+An+experiment+using+unsupervised+and+autosegmental+approach%2E+MSc+Thesis%2C+Addis+Ababa+University%2C+Ethiopia%2E)
10. Thorsten Brants. 2000. [TnT — a statistical part-of-speech tagger](https://md.research.elamur.ai/free/en/papers/brants-2000-tnt-statistical-part-speech-tagger/). In 6th Conf. Applied Natural Language Processing, pp. 224–231, Seattle, Wash. [↗](https://doi.org/10.48550/arxiv.cs/0003055)
11. Eric Brill and Jun Wu. 1998. Classifier combination for improved lexical disambiguation. In 17th Int. Conf. on Computational Linguistics, pp. 191–195, Montreal, Canada. [find](https://scholar.google.com/scholar?q=Eric+Brill+and+Jun+Wu%2E+1998%2E+Classifier+combination+for+improved+lexical+disambiguation%2E+In+17th+Int%2E+Conf%2E+on+Computational+Linguistics%2C+pp%2E+191%E2%80%93195%2C+Montreal%2C+Canada%2E)
12. Eric Brill. 1995. [Transformation-based error-driven learning and Natural Language Processing: A case study in part of speech tagging](https://md.research.elamur.ai/free/en/papers/brill-1995-transformation-based-error-driven-learning/). Computational Linguistics, 21:543–565.
13. CIA. 2009. The World Factbook — Ethiopia. The Central Intelligence Agency, Washington, DC. [Updated 22/01/09.] [find](https://scholar.google.com/scholar?q=CIA%2E+2009%2E+The+World+Factbook+%E2%80%94+Ethiopia%2E+The+Central+Intelligence+Agency%2C+Washington%2C+DC%2E+%5BUpdated+22%2F01%2F09%2E%5D)
14. Michael Collins. 2002. [Discriminative training methods for hidden Markov models: Theory and experiments with perceptron algorithms](https://md.research.elamur.ai/free/en/papers/collins-2002-discriminative-training-methods-hidden-markov/). In Empirical Methods in Natural Language Processing, pp. 1–8, Philadelphia, Penn.
15. Girma A. Demeke and Mesfin Getachew. 2006. Manual annotation of Amharic news items with part-of-speech tags and its challenges. ELRC Working Papers, 2:1–17. [find](https://scholar.google.com/scholar?q=Girma+A%2E+Demeke+and+Mesfin+Getachew%2E+2006%2E+Manual+annotation+of+Amharic+news+items+with+part%2Dof%2Dspeech+tags+and+its+challenges%2E+ELRC+Working+Papers%2C+2%3A1%E2%80%9317%2E)
16. Mona Diab, Kadri Hacioglu, and Daniel Jurafsky. 2004. [Automatic tagging of Arabic text: From raw text to base phrase chunks](https://md.research.elamur.ai/free/en/papers/diab-2004-automatic-tagging-arabic-text-raw/). In HLT Conf. North American ACL, pp. 149–152, Boston, Mass.
17. Thomas G. Dietterich. 1997. Machine-learning research: Four current directions. AI magazine, 18:97–136. [find](https://scholar.google.com/scholar?q=Thomas+G%2E+Dietterich%2E+1997%2E+Machine%2Dlearning+research%3A+Four+current+directions%2E+AI+magazine%2C+18%3A97%E2%80%93136%2E)
18. Kevin Duh and Katrin Kirchhoff. 2005. [POS tagging of dialectal Arabic: A minimally supervised approach](https://md.research.elamur.ai/free/en/papers/duh-2005-pos-tagging-dialectal-arabic-minimally/). Computational Approaches to Semitic Languages, pp. 55–62, Ann Arbor, Mich.
19. Sisay Fissaha and Johann Haller. 2003. Amharic verb lexicon in the context of machine translation. In 10th Traitement Automatique des Langues Naturelles, vol. 2, pp. 183–192, Batz-sur-Mer, France. [find](https://scholar.google.com/scholar?q=Sisay+Fissaha+and+Johann+Haller%2E+2003%2E+Amharic+verb+lexicon+in+the+context+of+machine+translation%2E+In+10th+Traitement+Automatique+des+Langues+Naturelles%2C+vol%2E+2%2C+pp%2E+183%E2%80%93192%2C+Batz%2Dsur%2DMer%2C+France%2E)
20. Sisay Fissaha. 2005. [Part of speech tagging for Amharic using conditional random fields](https://md.research.elamur.ai/free/en/papers/adafre-2005-part-speech-tagging-amharic-conditional/). Computational Approaches to Semitic Languages, pp. 47–54, Ann Arbor, Mich.
21. Mesfin Getachew. 2001. Automatic part of speech tagging for Amharic: An experiment using stochastic hidden Markov model (HMM) approach. MSc Thesis, Addis Ababa University, Ethiopia. [find](https://scholar.google.com/scholar?q=Mesfin+Getachew%2E+2001%2E+Automatic+part+of+speech+tagging+for+Amharic%3A+An+experiment+using+stochastic+hidden+Markov+model+%28HMM%29+approach%2E+MSc+Thesis%2C+Addis+Ababa+University%2C+Ethiopia%2E)
22. Jesús Giménez and Lluı́s Màrquez. 2004. [SVMTool: A general POS tagger generator based on support vector machines](https://md.research.elamur.ai/free/en/papers/gimenez-2004-svmtool-general-pos-tagger-generator/). In 4th Int. Conf. Language Resources and Evaluation, pp. 168–176, Lisbon, Portugal.
23. Sharon Goldwater and Thomas L. Griffiths. 2007. [A fully Bayesian approach to unsupervised part-of-speech tagging](https://md.research.elamur.ai/free/en/papers/goldwater-2007-fully-bayesian-approach-unsupervised-part/). In 45th ACL, pp. 744–751, Prague, Czech Rep.
24. Grover Hudson. 1999. Linguistic analysis of the 1994 Ethiopian census. Northeast African Studies, 6:89–107. [↗](https://doi.org/10.1353/nas.2003.0003)
25. Saib Mansour. 2008. Combining character and morpheme based models for part-of-speech tagging of Semitic languages. MSc Thesis, Technion, Haifa, Israel. [find](https://scholar.google.com/scholar?q=Saib+Mansour%2E+2008%2E+Combining+character+and+morpheme+based+models+for+part%2Dof%2Dspeech+tagging+of+Semitic+languages%2E+MSc+Thesis%2C+Technion%2C+Haifa%2C+Israel%2E)
26. Andrew Kachites McCallum. 2002. MALLET: A machine learning for language toolkit. Webpage. [find](https://scholar.google.com/scholar?q=Andrew+Kachites+McCallum%2E+2002%2E+MALLET%3A+A+machine+learning+for+language+toolkit%2E+Webpage%2E)
27. Guy De Pauw, Gilles-Maurice de Schryver, and Peter W. Wagacha. 2006. Data-driven part-of-speech tagging of Kiswahili. In 9th Int. Conf. Text, Speech and Dialogue, pp. 197–204, Brno, Czech Rep. [find](https://scholar.google.com/scholar?q=Guy+De+Pauw%2C+Gilles%2DMaurice+de+Schryver%2C+and+Peter+W%2E+Wagacha%2E+2006%2E+Data%2Ddriven+part%2Dof%2Dspeech+tagging+of+Kiswahili%2E+In+9th+Int%2E+Conf%2E+Text%2C+Speech+and+Dialogue%2C+pp%2E+197%E2%80%93204%2C+Brno%2C+Czech+Rep%2E)
28. Adwait Ratnaparkhi. 1996. [A maximum entropy model for part-of-speech tagging](https://md.research.elamur.ai/free/en/papers/ratnaparkhi-1996-maximum-entropy-model-part-speech/). In Empirical Methods in Natural Language Processing, pp. 133–142, Philadelphia, Penn.
29. Danny Shacham and Shuly Wintner. 2007. [Morphological disambiguation of Hebrew: A case study in classifier combination](https://md.research.elamur.ai/free/en/papers/shacham-2007-morphological-disambiguation-hebrew-case-study/). In Empirical Methods in Natural Language Processing, pp. 439–447, Prague, Czech Rep.
30. Drahomirá Spoustová, Jan Hajič, Jan Votrubec, Pavel Krbec, and Pavel Květoň. 2007. The best of two worlds: Cooperation of statistical and rule-based taggers for Czech. Balto-Slavonic Natural Language Processing, pp. 67–74. Prague, Czech Rep. [find](https://scholar.google.com/scholar?q=Drahomira%CC%81+Spoustova%CC%81%2C+Jan+Hajic%CC%8C%2C+Jan+Votrubec%2C+Pavel+Krbec%2C+and+Pavel+Kve%CC%8Cton%CC%8C%2E+2007%2E+The+best+of+two+worlds%3A+Cooperation+of+statistical+and+rule%2Dbased+taggers+for+Czech%2E+Balto%2DSlavonic+Natural+Lang)
31. Kristina Toutanova and Mark Johnson. 2007. A Bayesian LDA-based model for semi-supervised part-of-speech tagging. In 21st Int. Conf. Advances in Neural Information Processing Systems, pp. 1521–1528, Vancover, B.C. [find](https://scholar.google.com/scholar?q=Kristina+Toutanova+and+Mark+Johnson%2E+2007%2E+A+Bayesian+LDA%2Dbased+model+for+semi%2Dsupervised+part%2Dof%2Dspeech+tagging%2E+In+21st+Int%2E+Conf%2E+Advances+in+Neural+Information+Processing+Systems%2C+pp%2E+1521%E2%80%931528%2C+V)
32. Kristina Toutanova, Dan Klein, Christopher D. Manning, and Yoram Singer. 2003. [Feature-rich part-of-speech tagging with a cyclic dependency network](https://md.research.elamur.ai/free/en/papers/toutanova-2003-feature-rich-part-speech-tagging/). In HLT Conf. North American ACL, pp. 173–180, Edmonton, Alberta.
33. Atro Voutilainen. 1999. An experiment on the upper bound of interjudge agreement: The case of tagging. In 9th European ACL, pp. 204–208, Bergen, Norway. [find](https://scholar.google.com/scholar?q=Atro+Voutilainen%2E+1999%2E+An+experiment+on+the+upper+bound+of+interjudge+agreement%3A+The+case+of+tagging%2E+In+9th+European+ACL%2C+pp%2E+204%E2%80%93208%2C+Bergen%2C+Norway%2E)
34. Daniel Yacob. 1997. The System for Ethiopic Representation in ASCII — 1997 standard. Webpage. [find](https://scholar.google.com/scholar?q=Daniel+Yacob%2E+1997%2E+The+System+for+Ethiopic+Representation+in+ASCII+%E2%80%94+1997+standard%2E+Webpage%2E)

*Updated 1 November 2026*