---
title: "Identification of Languages in Algerian Arabic Multilingual"
description: "Adouane and Dobnik build a system that assigns a language label to every word… 93.14% accuracy. 14 references, cited by 12 papers here."
authors: ["Wafia Adouane", "Simon Dobnik"]
year: 2017
venue: "Proceedings of the Third Arabic Natural Language Processing Workshop"
source: "https://aclanthology.org/W17-1301/"
doi: "10.18653/v1/w17-1301"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/adouane-2017-identification-languages-algerian-arabic-multilingual/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/adouane-2017-identification-languages-algerian-arabic-multilingual/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2017](https://md.research.elamur.ai/free/en/papers/2017/page/4/) / Identification of Languages in Algerian Arabic Multilingual Documents

# Identification of Languages in Algerian Arabic Multilingual Documents

**Authors:** [Wafia Adouane](https://md.research.elamur.ai/free/en/authors/wafia-adouane/) and [Simon Dobnik](https://md.research.elamur.ai/free/en/authors/simon-dobnik/) · **Year:** 2017 · **Venue:** [Proceedings of the Third Arabic Natural Language Processing Workshop](https://md.research.elamur.ai/free/en/venues/wanlp/) ([WANLP 2017](https://md.research.elamur.ai/free/en/venues/wanlp/2017/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/W17-1301/) · [PDF](https://aclanthology.org/W17-1301.pdf) · [DOI 10.18653/v1/w17-1301](https://doi.org/10.18653/v1/w17-1301) · [OpenAlex](https://openalex.org/W2739911518)

**Citations:** 23 citations (OpenAlex, 2026-10-14) · [12](https://md.research.elamur.ai/free/en/papers/adouane-2017-identification-languages-algerian-arabic-multilingual/cited-by/) in our library, of the 149,439 papers whose bibliographies we read

## Summary
Social media posts by Algerian Arabic speakers often switch between several languages within one text, while most NLP tools assume a document is in a single language. Adouane and Dobnik build a system that assigns a language label to every word, using its context. They then group the labelled words into contiguous language chunks. They compiled a new corpus of 10,586 Algerian Arabic social media documents in Arabic script, labelled with nine classes. Annotation was sped up by iterative bootstrapping: an HMM tagger pre-labelled each new batch and humans corrected it. Separately compiled lexicons hold only words that belong unambiguously to one language. The final hybrid tagger chains several components:
- n-gram back-off taggers that can return an explicit UNK label,
- lexicon lookup,
- hand-written linguistic rules.

It reaches 93.14% accuracy against a 55.10% majority-class baseline. The work matters because it separates borrowing from code-switching, which earlier Arabic work did not do, and it gives a first such resource for an under-studied dialect.

## Key points
- The nine tag classes are ALG, MSA, FRC, BER, ENG, BOR (foreign words adapted to Algerian morphology), NER, SND (sounds and interjections) and DIG. Punctuation (PUN) and emoticons (EMO) are added later for chunking.
- Corpus and annotation:
  - 10,597 documents were collected and filtered to 10,586 Algerian Arabic documents (215,843 tokens).
  - Cohen's κ on the first 1,000 human-labelled documents was 89.27%.
  - Three annotators checked the whole bootstrapped corpus in about five weeks.
- The split is 10,008 training documents (215,832 tokens) and 578 evaluation documents (10,107 tokens). Evaluation documents were kept out of lexicon building.
- Results by stage:
  - The HMM tagger alone scores 85.88%. It has high precision but very poor recall on rare classes, for example BER recall of 18.18%.
  - BackOff(Trigram, Bigram, Unigram, UNK) plus lexicons scores 92.86%.
  - Adding rules raises this to 93.14%.
  - Cross-validation gives similar figures.
- 476 tokens remain UNK because they appear neither in training data nor in the lexicons. These are flagged as material for growing the lexicons.

## Interesting ideas
- Instead of letting a smoothed model quietly guess a label for unseen words, the system emits an explicit UNK class, which avoids having to pick arbitrary confidence thresholds.
- The corpus was annotated in growing batches: a model trained on everything checked so far pre-labels the next batch and humans correct it, which saves effort and exposes the model's typical errors along the way.
- Lexicons are cleaned so they contain only words unambiguous to one class, treating the dictionary as a store of exceptions while ambiguous items go to context-based disambiguation.
- The components form a fall-back chain (context n-grams, then back-off, then lexicon lookup, then rules), each catching what the previous one left unresolved, so every label can be traced to the component that produced it.
- The UNK residue is treated as a work queue for extending coverage, not just as an error count.

## For Mova

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

- **I1** (tokens): The en tagger and lemmatizer could output an honest "unknown" for out-of-vocabulary forms rather than a forced guess, so later stages and explanations can see exactly where knowledge ran out.
- **I2** (learn): Silver data from witness parsers could be released in rounds, with each corrected round retraining the parsers before the next one, and the corrections logged as error statistics.
- **I3** (world): Compiled knowledge such as verb classes or ontology mappings could separate single-reading entries, applied directly, from ambiguous ones that must be resolved from context.
- **I4** (tokens): A tagging cascade that records which layer decided each tag would fit Mova's explainability requirement and show which layer to improve.
- **I5** (learn): Tokens or constructions that Mova cannot handle could be collected automatically and sent to LLM teacher tiers for labelling, closing a coverage-growth loop.

*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

- [University of Gothenburg](https://md.research.elamur.ai/free/en/universities/university-of-gothenburg/) — University, Gothenburg, Sweden · 34 works · 1,625 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 · [🌍 Multilingual & cross-lingual NLP](https://md.research.elamur.ai/free/en/topics/multilingual-nlp/) · 828 works

## Cite

Cite as: Wafia Adouane and Simon Dobnik. 2017. Identification of Languages in Algerian Arabic Multilingual Documents. Proceedings of the Third Arabic Natural Language Processing Workshop.

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

```bibtex
@inproceedings{adouane2017identification,
  title = {{Identification of Languages in Algerian Arabic Multilingual Documents}},
  author = {Wafia Adouane and Simon Dobnik},
  booktitle = {Proceedings of the Third Arabic Natural Language Processing Workshop},
  year = {2017},
  doi = {10.18653/v1/w17-1301},
  url = {https://doi.org/10.18653/v1/w17-1301}
}
```

- [← previous SiTAKA at SemEval-2017 Task 4: Sentiment Analysis in Twitter Based on a Rich Set of Features](https://md.research.elamur.ai/free/en/papers/jabreel-2017-sitaka-semeval-2017-task-4/)
- [next → Ontology-Aware Token Embeddings for Prepositional Phrase Attachment](https://md.research.elamur.ai/free/en/papers/dasigi-2017-ontology-aware-token-embeddings-prepositional/)
- [↑ back to papers of 2017](https://md.research.elamur.ai/free/en/papers/2017/page/4/)

## References

14 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Wafia Adouane, Nasredine Semmar, Richard Johansson, and Victoria Bobicev. 2016. [Automatic detection of Arabicized Berber and Arabic varieties](https://md.research.elamur.ai/free/en/papers/adouane-2016-automatic-detection-arabicized-berber-arabic/). In Proceedings of the Third Workshop on NLP for Similar Languages, Varieties and Dialects, pages 63–72.
2. Jean Carletta. 1996. [Assessing agreement on classification tasks: the kappa statistic](https://md.research.elamur.ai/free/en/papers/carletta-1996-assessing-agreement-classification-tasks-kappa/). Computational Linguistics, 2(22):249–254. [↗](https://doi.org/10.48550/arxiv.cmp-lg/9602004)
3. Marine Carpuat. 2014. Mixed-language and code-switching in the canadian hansard. In Proceedings of The First Workshop on Computational Approaches to Code Switching, pages 107–115. [find](https://scholar.google.com/scholar?q=Marine+Carpuat%2E+2014%2E+Mixed%2Dlanguage+and+code%2Dswitching+in+the+canadian+hansard%2E+In+Proceedings+of+The+First+Workshop+on+Computational+Approaches+to+Code+Switching%2C+pages+107%E2%80%93115%2E)
4. Amitava Das and Björn Gambäck. 2013. Code-mixing in social media text: The last language identification frontier? Traitement Automatique des Langues, 54(3):41–64. [find](https://scholar.google.com/scholar?q=Amitava+Das+and+Bj%C3%B6rn+Gamb%C3%A4ck%2E+2013%2E+Code%2Dmixing+in+social+media+text%3A+The+last+language+identification+frontier%3F+Traitement+Automatique+des+Langues%2C+54%283%29%3A41%E2%80%9364%2E)
5. Heba Elfardy and Mona Diab. 2012. Token level identification of linguistic code switching. In COLING, pages 287–296. Conclusions and Future Work We have presented a system for identifying the language at word and long sequence levels in multilingual documents in Algerian Arabic. We de-Heba Elfardy, Mohamed Al-Badrashiny, and Mona Diab. 2013. Code switch point detection in arabic. In Proceedings of the 18th International Conference on Application of Natural Language to Information Systems (NLDB2013). [find](https://scholar.google.com/scholar?q=Heba+Elfardy+and+Mona+Diab%2E+2012%2E+Token+level+identification+of+linguistic+code+switching%2E+In+COLING%2C+pages+287%E2%80%93296%2E+Conclusions+and+Future+Work+We+have+presented+a+system+for+identifying+the+language)
6. Dong Nguyen and A. Seza Doğruöz. 2013. Word level language identification in online multilingual communication. In Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 18–21. [find](https://scholar.google.com/scholar?q=Dong+Nguyen+and+A%2E+Seza+Dog%CC%86ru%C3%B6z%2E+2013%2E+Word+level+language+identification+in+online+multilingual+communication%2E+In+Proceedings+of+the+2013+Conference+on+Empirical+Methods+in+Natural+Language+Processi)
7. Mario Piergallini, Rouzbeh Shirvani, Gauri S. Gautam, and Mohamed Chouikha. 2016. Word-level language identification and predicting code-switching points in swahili-english language data. In Proceedings of the Second Workshop on Computational Approaches to Code Switching, pages 21–29. [find](https://scholar.google.com/scholar?q=Mario+Piergallini%2C+Rouzbeh+Shirvani%2C+Gauri+S%2E+Gautam%2C+and+Mohamed+Chouikha%2E+2016%2E+Word%2Dlevel+language+identification+and+predicting+code%2Dswitching+points+in+swahili%2Denglish+language+data%2E+In+Proceedin)
8. Shana Poplack and Marjory Meechan. 1998. How languages fit together in codemixing. The International Journal of Bilingualism, 2(2):127–138. [find](https://scholar.google.com/scholar?q=Shana+Poplack+and+Marjory+Meechan%2E+1998%2E+How+languages+fit+together+in+codemixing%2E+The+International+Journal+of+Bilingualism%2C+2%282%29%3A127%E2%80%93138%2E)
9. Younes Samih and Wolfgang Maier. 2016. Detecting code-switching in moroccan arabic. In Proceedings of SocialNLP @ IJCAI-2016. [find](https://scholar.google.com/scholar?q=Younes+Samih+and+Wolfgang+Maier%2E+2016%2E+Detecting+code%2Dswitching+in+moroccan+arabic%2E+In+Proceedings+of+SocialNLP+%40+IJCAI%2D2016%2E)
10. Burr Settles. 2009. [Active learning literature survey](https://md.research.elamur.ai/free/en/papers/settles-2009-active-learning-literature-survey/). Computer Sciences Technical Report 1648, University of Wisconsin–Madison.
11. Anil Kumar Singh and Jagadeesh Gorla. 2007. Identification of languages and encodings in a multilingual document. In Building and Exploring Web Corpora (WAC3-2007): Proceedings of the 3rd Web as Corpus Workshop, Incorporating Cleaneval. [find](https://scholar.google.com/scholar?q=Anil+Kumar+Singh+and+Jagadeesh+Gorla%2E+2007%2E+Identification+of+languages+and+encodings+in+a+multilingual+document%2E+In+Building+and+Exploring+Web+Corpora+%28WAC3%2D2007%29%3A+Proceedings+of+the+3rd+Web+as+Corpu)
12. Thamar Solorio, Elizabeth Blair, Suraj Maharjan, Steven Bethard, Mona Diab, Mahmoud Ghoneim, Abdelati Hawwari, Fahad AlGhamdi, Julia Hirschberg, Alison Chang, and Pascale Fung. 2014. [Overview for the first shared task on language identification in code-switched data](https://md.research.elamur.ai/free/en/papers/solorio-2014-overview-first-shared-task-language/). In Proceedings of The First Workshop on Computational Approaches to Code Switching, pages 62–72.
13. Lameen Souag. 2000. Grammar of Algerian Darja. Retrieved from [https://goo.gl/p2EmVH](https://goo.gl/p2EmVH). [↗](https://goo.gl/p2EmVH)
14. Bird Steven, Ewan Klein, and Edward Loper. 2009. Natural Language Processing with Python. O’Reilly Media. [find](https://scholar.google.com/scholar?q=Bird+Steven%2C+Ewan+Klein%2C+and+Edward+Loper%2E+2009%2E+Natural+Language+Processing+with+Python%2E+O%E2%80%99Reilly+Media%2E)

*Updated 1 November 2026*