Elamur Research⌕
☰
home / papers / 2018 / Part-of-Speech Tagging for Arabic Gulf Dialect Using Bi-LSTM

Part-of-Speech Tagging for Arabic Gulf Dialect Using Bi-LSTM

Paper · 2018 · Randah Alharbi, Walid Magdy, Kareem Darwish et al.

Summary

Standard Arabic taggers break down on dialectal social-media text, and Gulf Arabic had no dedicated tagger at all, although it is widely used online. The authors built and released a POS-annotated Gulf Arabic dataset with annotation guidelines. They measured how far an MSA tool falls short and then trained two dialect-specific taggers: an SVM classifier and a Bi-LSTM sequence labeler. Both work on clitic-level units. The authors added hand-designed features: a "meta-type" signal for tokens such as mentions, hashtags, numbers and URLs, and a stem-template (morphological pattern) feature. They also compared word, character and combined input representations and tested an enlarged training set (Gulf++). The MSA tagger Farasa reaches about 75% on Gulf text. The best Bi-LSTM, which combines character and word representations with both features, reaches 91.2%. The best SVM reaches about 86%. The study shows that a dedicated dialect tool pays off substantially compared with reusing MSA tools, and that a neural model copes better with unstandardised spelling than feature-heavy classifiers.

Key points

  • Baseline: the Farasa MSA tagger scores 75% on Gulf Arabic. The best Bi-LSTM (CC2W+W representation, meta-type and template features, Gulf++ data) scores 91.2%.
  • For the best system, out-of-vocabulary accuracy is 73.5%, and precision, recall and F-score are 83.9%, 91.2% and 87.4%.
  • The best SVM reaches 85.96%, using TF-IDF clitic values plus meta-type features. Bi-LSTM beats SVM in most configurations.
  • The most frequent confusions are verb→noun (16.6%), adjective→noun (16.0%) and pronoun→noun-suffix (12.4%). The template feature lifted adjective accuracy from 67.4% to 71.6%.
  • The Gulf POS dataset and its guidelines are freely available via QCRI's DA resources page.

Interesting ideas

  • I1. Before building a dedicated tool, first measure how badly the off-the-shelf standard-variety tagger does on the variety, and analyse its failures; this turns a hunch into a quantified gap that justifies the work.
  • I2. A small set of deterministic "meta-type" signals for tokens like mentions, hashtags, numbers and URLs made those tags essentially perfect (100%), which frees the model to focus on genuinely ambiguous words.
  • I3. A morphological-template feature (the shape of the stem) helped separate adjectives from nouns, because the two classes follow consistent patterns; abstracting word shape generalises beyond seen vocabulary.
  • I4. The authors found that a neural sequence model beat feature-engineered SVMs where spelling and grammar are unstandardised, since hand features presuppose norms the data lacks.
  • I5. Combining character-level and word-level representations worked best once clitic-level features were added, while plain word representations did better with little training data.

For Mova

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

  • I1 (tokens): Run the English tagger on informal or dialectal registers (tweets, chat, transcribed speech) and report the drop per tag before deciding whether register-specific training is needed.
  • I2 (tokens): Add explicit, explainable type flags to the tokenizer output so the tagger never wastes capacity on mechanically recognisable tokens.
  • I3 (tokens): Suffix and shape templates (e.g. -ous, -ly, -tion) could serve as transparent features to reduce adjective/noun and verb/noun confusions on unknown words.
  • I4 (learn): When training on noisy silver data from non-standard text, prefer letting the learned model absorb variation, keeping hand features only where they encode stable regularities.
  • I5 (tree): For gparse, pick input representation by data size, starting with word-level inputs for small dialect treebanks and adding character composition as data grows.

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 EdinburghEdinburgh · United Kingdom381 works · 80,747 citations🎓Hamad bin Khalifa UniversityDoha · Qatar144 works · 14,175 citations🎓Umm al-Qura UniversityMakkah al Mukarramah · Saudi Arabia21 works · 274 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 works

🧭 Topics 2Research subjects and languages the work is about.Semitic languages · 3,066 worksPart-of-speech tagging · 126 works

Cite

Cite as Randah Alharbi, Walid Magdy, Kareem Darwish, Ahmed AbdelAli and Hamdy Mubarak. 2018. Part-of-Speech Tagging for Arabic Gulf Dialect Using Bi-LSTM. Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018).

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

@inproceedings{alharbi2018part,
  title = {{Part-of-Speech Tagging for Arabic Gulf Dialect Using Bi-LSTM}},
  author = {Randah Alharbi, Walid Magdy, Kareem Darwish, Ahmed AbdelAli and Hamdy Mubarak},
  booktitle = {Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018)},
  year = {2018},
  doi = {10.63317/4bgx23eisvz7},
  url = {https://doi.org/10.63317/4bgx23eisvz7}
}

References

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

  1. Abdelali, A., Darwish, K., Durrani, N., and Mubarak, H. (2016). Farasa: A Fast and Furious Segmenter for Arabic. Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Demonstrations, 2016:11–16.
  2. Abuata, B. and Al-Omari, A. (2015). A rule-based stemmer for Arabic Gulf dialect. Journal of King Saud University - Computer and Information Sciences, 27(2):104– 112. find
  3. Al-sabbagh, R. and Girju, R. (2012). A Supervised POS Tagger for Written Arabic Social Networking Corpora. Proceedings of KONVENS 2012 (Main track: oral presentations), Vienna, September 19, 2012:39–52. find
  4. Boudlal, A., Belahbib, R., Lakhouaja, A., Mazroui, A., Meziane, A., and Bebah, M. (2011). A markovian approach for Arabic root extraction. International Arab Journal of Information Technology, 8(1):91–98. find
  5. Craig Hagerman. (2012). Evaluating the Performance of Automated Part-of-Speech Taggers on an L2 Corpus. pages 29–42. find
  6. Darwish, K. and Magdy, W. (2014). Arabic Information Retrieval. Foundations and Trends R in Information Retrieval, 7(4):239–342. find
  7. Darwish, K., Magdy, W., and Mourad, A. (2012). Language processing for arabic microblog retrieval. In Proceedings of the 21st ACM international conference on Information and knowledge management, pages 2427– 2430. ACM. find
  8. Darwish, K., Abdelali, A., and Mubarak, H. (2014). Using Stem-Templates to improve Arabic POS and Gender/Number Tagging. International Conference on Language Resources and Evaluation (LREC-2014), pages 2926–2931.
  9. Darwish, K., Mubarak, H., and Abdelali, A. (2017). Arabic POS Tagging : Don ’ t Abandon Feature Engineering Just Yet. WANLP 2017 (co-located with EACL 2017), pages 130–137.
  10. Diab, M. and Habash, N. (2007). Arabic dialect processing tutorial. Proceedings of the Human Language Technology Conference of the NAACL, Companion Volume: Tutorial Abstracts on XX - NAACL ’07, (April):5–6. find
  11. Diab, M., Habash, N., Rambow, O., Altantawy, M., and Benajiba, Y. (2010). COLABA: Arabic dialect annotation and processing. LREC Workshop on Semitic Language Processing, (January 2016):66–74. find
  12. Diab, M. (2009). Second Generation AMIRA Tools for Arabic Processing : Fast and Robust Tokenization , POS tagging , and Base Phrase Chunking. Proceedings of the Second International Conference on Arabic Language Resources and Tools, pages 285–288. find
  13. Duh, K. and Kirchhoff, K. (2005). POS Tagging of Dialectal Arabic : A Minimally Supervised Approach. Computational Linguistics, (June):55–62.
  14. Giménez, J. and Màrquez, L. (2003). Fast and accurate part-of-speech tagging: The SVM approach revisited. Ranlp, pages 153–163. find
  15. Habash, N. and Rambow, O. (2006). MAGEAD: a morphological analyzer and generator for the Arabic dialects. Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the Association for Computational Linguistics (Coling-ACL), M:681–688.
  16. Habash, N., Diab, M., and Rambow, O. (2012a). Conventional Orthography for Dialectal Arabic. Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC-2012), (January 2012):711–718.
  17. Habash, N., Eskander, R., and Hawwari, A. (2012b). A Morphological Analyzer for Egyptian Arabic. Proceedings of the Twelfth Meeting of the Special Interest Group on Computational Morphology and Phonology SIGMORPHON2012, pages 1–9.
  18. Habash, N. Y. (2010). Introduction to Arabic Natural Language Processing, volume 3. Morgan & Claypool Pu-plisher. find
  19. Joachims, T. (2002). Learning to Classify Text Using Support Vector Machines: Methods, Theory and Algorithms, volume 29. SPRINGER SCIENCE+BUSINESS ME-DIA. LLC. find
  20. Joachims, T. (2008). SVM-Multiclass: Multi-Class Support Vector Machine. find
  21. Jurafsky, D. and Martin, J. H. (2009). Speech and Language Processing: An Introduction to Natural Language Processing, Computational Linguistics, and Speech Recognition. Speech and Language Processing An Introduction to Natural Language Processing Computational Linguistics and Speech Recognition, 21:0–934. find
  22. Katz, G. and Diab, M. (2011). Introduction to the Special Issue on Arabic Computational Linguistics. ACM Transactions on Asian Language Information Processing, 10(1):1–4. find
  23. Khalifa, S., Habash, N., Abdulrahim, D., and Hassan, S. (2016). A Large Scale Corpus of Gulf Arabic. Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC 2016), pages 4282– 4289.
  24. Khoja, S. (2001). APT : Arabic Part-Of-speech Tagger. Proceedings of the Student Workshop at NAACL, pages 20—-25. find
  25. Ling, W., Luis, T., Marujo, L. L., Astudillo, R. F., Amir, S., Dyer, C., Black, A. W., Trancoso, I., Fermandez, R., Amir, S., Marujo, L. L., and Luı́s, T. (2015). Finding Function in Form: Compositional Character Models for Open Vocabulary Word Representation. Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, (September):1520–1530.
  26. Manning, C. D., Ragahvan, P., and Schutze, H. (2009). An Introduction to Information Retrieval. Information Retrieval, (c):1–18. find
  27. Pasha, A., Al-badrashiny, M., Diab, M., Kholy, A. E., Eskander, R., Habash, N., Pooleery, M., Rambow, O., and Roth, R. M. (2014). MADAMIRA : A Fast , Comprehensive Tool for Morphological Analysis and Disambiguation of Arabic. Proceedings of the 9th Language Resources and Evaluation Conference (LREC’14), pages 1094–1101.
  28. Plank, B., Søgaard, A., and Goldberg, Y. (2016). Multilingual Part-of-Speech Tagging with Bidirectional Long Short-Term Memory Models and Auxiliary Loss. Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (ACL-16), pages 412– 418.
  29. Rambow, O., Habash, N., Rambow, O., and Roth, R. (2009). MADA + TOKAN : A toolkit for Arabic tokenization , diacritization , morphological disambiguation , POS Tagging, Stemming and Lemmatization. (January 2017). find
  30. Samih, Y., Eldesouki, M., Attia, M., Darwish, K., Abdelali, A., Mubarak, H., and Kallmeyer, L. (2017). Learning from Relatives : Unified Dialectal Arabic Segmentation. (CoNLL):1–10.
  31. Wang, P., Qian, Y., Soong, F. K., He, L., and Zhao, H. (2015). A Unified Tagging Solution: Bidirectional LSTM Recurrent Neural Network with Word Embedding. arxiv: EMNLP-reject. find
  32. Wilks, Y. (1996). Natural language processing. Commun. ACM, 39(1):60–62. find
  33. Zeroual, I., Lakhouaja, A., and Belahbib, R. (2017). Towards a standard Part of Speech tagset for the Arabic language. Journal of King Saud University - Computer and Information Sciences, 29(2):171–178. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/alharbi-2018-part-speech-tagging-arabic-gulf/