Elamur Research⌕
☰
home / papers / 2021 / Similarities between Arabic Dialects: Investigating Geographical Proximity

Similarities between Arabic Dialects: Investigating Geographical Proximity

Summary

Most Arabic dialect identification work assumes that a dialect corresponds to a region or a country. Newer corpora push this down to cities and provinces. The authors ask a related question: does the physical distance between two places predict how alike their written dialects are, and do national borders really matter?

They use two datasets:

  • MADAR, a parallel travel-domain corpus covering 25 cities.
  • The first release of NADI, Twitter data labelled with 100 provinces.

For each location they build a text profile and compare every pair of locations with cosine similarity. They also compute geographic distance from map coordinates. They then test how similarity relates to distance, both per location and across each dataset, and average similarity within widening distance bands.

Similarity falls steadily as distance grows. Some city pairs from different countries turn out more alike than the typical same-country pair. The authors conclude that country labels are a coarse proxy. Dialect identification should be framed and evaluated with geography in mind, for example by not scoring a neighbouring-city guess as a flat error.

Key points

  • Data: MADAR (25 cities, 300 location pairs) and NADI first release (100 provinces, about 4,939 pairs with distances). The method combines cosine similarity of texts with geographic distance, organised around six research questions on correlation and distance-band subsets.
  • MADAR: the 8 pairs closer than 200 km average 73.1% similarity, compared with 24% across all 300 pairs.
  • NADI: the 48 pairs closer than 100 km average 21.9% similarity, compared with 13.2% overall. Averages shrink monotonically as the distance cut-off widens.
  • Distance and similarity are significantly correlated, both for individual locations and for each dataset as a whole. Some cross-border pairs exceed the average similarity of same-country pairs.
  • The authors list several limitations:
    • NADI includes some non-Arabic text and partly overlaps with Twitter geolocation.
    • MADAR is travel-only and consists of commissioned translations rather than natural text.
    • Single coordinates for large provinces may misplace where texts come from.
    • Only one similarity measure was tried.

Interesting ideas

  • I1. Score near-misses with "tolerance distances", giving partial credit when a wrong prediction is close to the truth, instead of a binary right/wrong. — For Mova (learn): Evaluation could give graded credit when a parser picks a label or head that is "close" (e.g., obl vs nmod, or a sibling verb class), so error reports separate near-misses from real failures.
  • I2. Plot average similarity over cumulative distance bands (<100, <200, … km) to show a trend more clearly than a single correlation coefficient. — For Mova (tree): When converting between UD dialects or treebank genres, plot parser agreement or LAS against a cumulative distance measure between sources, to see where conversion or transfer starts to degrade.
  • I3. Test whether a grouping is justified by checking if cross-group pairs beat the within-group average similarity. — For Mova (world): Apply the same check to VerbNet classes or WordNet-derived ontology groups. Members that resemble another class more than their own become candidates for reassignment.
  • I4. Prefer finer units over coarse labels when the coarse label hides real variation, as cities do compared with countries. — For Mova (frames): Speech-act or irony categories could be split into finer subtypes when evidence shows the broad category mixes dissimilar cases, keeping explanations more precise.
  • I5. State openly that the choice of reference point (a province's coordinates) can bias the measured result. — For Mova (learn): Negative-control reports should record which anchors or reference sets a metric depends on, so a result is not mistaken for a property of the data.

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

Cited by

7 papers of our library cite it (of the 140884 whose bibliographies we read); 3 of them have a page here, 4 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.1016/j.ipm.2021.102770. · arXiv 2105.04221

@misc{alsudais2021similarities,
  title = {{Similarities between Arabic Dialects: Investigating Geographical Proximity}},
  author = {Alsudais, Abdulkareem and Alotaibi, Wafa and Alomary, Faye},
  year = {2021},
  doi = {10.1016/j.ipm.2021.102770.},
  eprint = {2105.04221},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.1016/j.ipm.2021.102770.}
}

References

43 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. M. Abdul-mageed, Modeling Arabic subjectivity and sentiment in lexical space, Inf. Process. https://doi.org/10.1016/j.ipm.2017.07.004. ↗
  2. I.A. Farha, W. Magdy, A comparative study of effective approaches for Arabic sentiment analysis, Inf. https://doi.org/10.1016/j.ipm.2020.102438. ↗
  3. S. Harrat, K. Meftouh, K. Smaili, Machine translation for Arabic dialects (survey), Inf. Process. https://doi.org/10.1016/j.ipm.2017.08.003. ↗
  4. K. Darwish, N. Habash, M. Abbas, H. Al-Khalifa, H.T. Al-Natsheh, H. Bouamor, K. Bouzoubaa, V. Cavalli-Sforza, S.R. El-Beltagy, W. El-Hajj, M. Jarrar, H. Mubarak, A Panoramic Survey of Natural Language Processing in Manag. (2019) 291–307. Process. Manag. (2021) 102438. Manag. (2019) 262–273. the Arab World, https://doi.org/10.1145/3447735. ↗
  5. M. Abdul-Mageed, C. Zhang, H. Bouamor, N. Habash, NADI 2020: The First Nuanced Arabic Dialect Identification Shared Task, in: Proc. Fifth Arab. Nat. Lang. Process. Work., Association for Computational Linguistics, Barcelona, Spain (Online), https://www.aclweb.org/anthology/2020.wanlp-1.9. ↗
  6. M. Abdul-Mageed, C. Zhang, A. Elmadany, H. Bouamor, N. Habash, NADI 2021: The Second Nuanced Arabic Dialect Identification Shared Task, in: Proc. Sixth Arab. Nat. Lang. Process. Work. (WANLP 2021), 2021. find
  7. A. Alshutayri, E. Atwell, Exploring Twitter as a source of an Arabic dialect corpus, Int. J. Comput. Linguist. 8 (2017) 37–44. find
  8. F. Sadat, F. Kazemi, A. Farzindar, Automatic Identification of Arabic Dialects in Social Media, in: Proc. First Int. Work. Soc. Media Retr. Anal., Association for Computing Machinery, New York, NY, USA, 2014: pp. 35–40. https://doi.org/10.1145/2632188.2632207. ↗
  9. Y. Khalifa, A. Elnagar, Colloquial Arabic Tweets: Collection, Automatic Annotation, and Classification, in: 2020 Int. Conf. Asian Lang. Process., 2020: pp. 163–168. https://doi.org/10.1109/IALP51396.2020.9310507. ↗
  10. A. Abdelali, H. Mubarak, Y. Samih, S. Hassan, K. Darwish, QADI: Arabic Dialect Identification in the Wild, in: Proc. Sixth Arab. Nat. Lang. Process. Work., Association for Computational Linguistics, Kyiv, Ukraine (Virtual), 2021: pp. 1–10. https://www.aclweb.org/anthology/2021.wanlp-1.1. ↗ our card
  11. S. Al-hanouf, S. Alshalan, R. Alshalan, N. Alrumayyan, S. Al-manea, N. Al-twairesh, R. Al-matham, N. Madi, N. Almugren, S. Bawazeer, N. Al-mutlaq, N. Almanea, W. Bin Huwaymil, S. Alshalan, R. Alshalan, N. Alrumayyan, S. Al-manea, R. Alotaibi, S. Al-senaydi, A. Alfutamani, S. Bawazeer, N. Al-Commun. ACM. (2021) 72–81. 2020: pp. 97–110. mutlaq, N. Almanea, W. Bin Huwaymil, D. Alqusair, R. Alotaibi, S. Al-Senaydi, A. Alfutamani, SUAR: Towards Building a Corpus for the Saudi Dialect, Procedia Comput. https://doi.org/10.1016/j.procs.2018.10.462. ↗
  12. H. Bouamor, N. Habash, M. Salameh, W. Zaghouani, O. Rambow, D. Abdulrahim, O. Obeid, S. Khalifa, F. Eryani, A. Erdmann, K. Oflazer, The MADAR Arabic Dialect Corpus and Lexicon, in: LREC, 2018. find
  13. K. Abu Kwaik, M. Saad, S. Chatzikyriakidis, S. Dobnik, Shami: A Corpus of Levantine Arabic Dialects, in: Proc. Elev. Int. Conf. Lang. Resour. Eval. 2018, European Language Resources Association (ELRA), Miyazaki, Japan, 2018. find
  14. R. Ziedan, M. Micheal, A. Alsammak, M. Mursi, A. Elmaghraby, A unified approach for arabic language dialect detection, in: 29th Int. Conf. Comput. Appl. Ind. Eng. (CAINE 2016), Denver, USA, 2016. find
  15. A. Alshutayri, E. Atwell, Arabic dialects annotation using an online game, in: 2018 2nd Int. Conf. Nat. Lang. Speech Process., 2018: pp. 1–5. https://doi.org/10.1109/ICNLSP.2018.8374371. ↗
  16. K. Almeman, Automatically Building VoIP Speech Parallel Corpora for Arabic Dialects, ACM Trans. Asian Low-Resour. Lang. Inf. Process. 17 (2017). https://doi.org/10.1145/3132708. ↗
  17. R. Zarnoufi, H. Jaafar, M. Abik, Machine Normalization: Bringing Social Media Text from Non-Standard to Standard Form, ACM Trans. Asian Low-Resour. Lang. Inf. Process. 19 (2020). https://doi.org/10.1145/3378414. ↗
  18. R. Tachicart, K. Bouzoubaa, S.L. Aouragh, H. Jaafa, Automatic identification of Moroccan colloquial Arabic, in: Int. Conf. Arab. Lang. Process., 2017: pp. 201–214. find
  19. M. Lichouri, M. Abbas, A.A. Freihat, D.E.H. Megtouf, Word-Level vs Sentence-Level Language Identification: Application to Algerian and Arabic Dialects, Procedia Comput. https://doi.org/https://doi.org/10.1016/j.procs.2018.10.484. Sci. (2018) 72–82. Sci. (2018) 246–253. ↗
  20. A. Masmoudi, S. Mdhaffar, R. Sellami, L.H. Belguith, Automatic Diacritics Restoration for Tunisian Dialect, ACM Trans. Asian Low-Resour. Lang. Inf. Process. 18 (2019). https://doi.org/10.1145/3297278. ↗
  21. A. Alnawas, N. Arici, Sentiment Analysis of Iraqi Arabic Dialect on Facebook Based on Distributed Representations of Documents, ACM Trans. Asian Low-Resour. Lang. Inf. Process. 18 (2019). https://doi.org/10.1145/3278605. ↗
  22. M. Al-ayyoub, A. Allah, Y. Jararweh, M.N. Al-kabi, M. Sa, A comprehensive survey of arabic sentiment analysis, Inf. Process. Manag. 56 (2019) 320–342. https://doi.org/10.1016/j.ipm.2018.07.006. ↗
  23. W. Farhan, B. Talafha, A. Abuammar, R. Jaikat, M. Al-ayyoub, A.B. Tarakji, A. Toma, R. Samsung, Unsupervised dialectal neural machine translation, Inf. Process. Manag. https://doi.org/10.1016/j.ipm.2019.102181. ↗
  24. A. Soufan, Deep Learning for Sentiment Analysis of Arabic Text, in: Proc. Arab. 6th Annu. Int. Conf. Res. Track, Association for Computing Machinery, New York, NY, USA, 2019. https://doi.org/10.1145/3333165.3333185. ↗
  25. Q. Zhang, J.H.L. Hansen, Language/Dialect Recognition Based on Unsupervised Deep Learning, IEEE/ACM Trans. Audio, Speech and Lang. Proc. 26 (2018) 873–882. https://doi.org/10.1109/TASLP.2018.2797420. ↗
  26. M. Alali, N. Mohd Sharef, M.A. Azmi Murad, H. Hamdan, N.A. Husin, Narrow Convolutional Neural Network for Arabic Dialects Polarity Classification, IEEE Access. https://doi.org/10.1109/ACCESS.2019.2929208. ↗
  27. M. Abdul-Mageed, C. Zhang, A. Elmadany, L. Ungar, Toward micro-dialect identification in diaglossic and code-switched environments, Proc. 2020 Conf. Empir. Methods Nat. Lang. Process. (2020). find
  28. O. Obeid, M. Salameh, H. Bouamor, N. Habash, ADIDA: Automatic Dialect Identification for Arabic, in: Proc. 2019 Conf. North {A}merican Chapter Assoc. Comput. Linguist., Association for Computational Linguistics, (2020) 102181. (2019) 96272–96283. Minneapolis, Minnesota, 2019: pp. 6–11. https://doi.org/10.18653/v1/N19-4002. ↗
  29. C. Zhang, M. Abdul-Mageed, No Army, No Navy: BERT Semi-Supervised Learning of Arabic Dialects, in: Proc. Fourth Arab. Nat. Lang. Process. Work., Association for Computational Linguistics, Florence, Italy, 2019: pp. 279–284. https://doi.org/10.18653/v1/W19-4637. ↗
  30. P. Mishra, V. Mujadia, Arabic dialect identification for travel and twitter text, in: Proc. Fourth Arab. Nat. Lang. Process. Work., 2019: pp. 234–238. find
  31. A. Ragab, H. Seelawi, M. Samir, A. Mattar, H. Al-Bataineh, M. Zaghloul, A. Mustafa, B. Talafha, A.A. Freihat, H. Al-Natsheh, Mawdoo3 ai at madar shared task: Arabic fine-grained dialect identification with ensemble learning, in: Proc. Fourth Arab. Nat. Lang. Process. Work., 2019: pp. 244–248. find
  32. B. Talafha, M. Ali, M.E. Za’ter, H. Seelawi, I. Tuffaha, M. Samir, W. Farhan, H. Al-Natsheh, Multi-dialect Arabic BERT for Country-level Dialect Identification, in: Proc. Fifth Arab. Nat. Lang. Process. Work., Association for Computational Linguistics, Barcelona, Spain (Online), 2020: pp. 111–118. https://www.aclweb.org/anthology/2020.wanlp-1.10. ↗
  33. A. El Mekki, A. Alami, H. Alami, A. Khoumsi, I. Berrada, Weighted combination of BERT and N-GRAM features for Nuanced Arabic Dialect Identification, in: Proc. Fifth Arab. Nat. Lang. Process. Work., Association for Computational Linguistics, Barcelona, Spain (Online), 2020: pp. 268–274. find
  34. N. AlShenaifi, A. Azmi, Faheem at NADI shared task: Identifying the dialect of Arabic tweet, in: Proc. Fifth Arab. Nat. Lang. Process. Work., Association for Computational Linguistics, Barcelona, Spain (Online), 2020: pp. 282–287. https://www.aclweb.org/anthology/2020.wanlp-1.29. ↗
  35. K.A. Kwaik, M. Saad, S. Chatzikyriakidis, S. Dobnika, A Lexical Distance Study of Arabic Dialects, Procedia Comput. Sci. 142 (2018) 2–13. https://doi.org/https://doi.org/10.1016/j.procs.2018.10.456. ↗
  36. H. Bouamor, N. Habash, K. Oflazer, A Multidialectal Parallel Corpus of Arabic., in: LREC, 2014: pp. 1240–1245. find our card
  37. S. Bird, E. Klein, E. Loper, Natural Language Processing with Python, 1st ed., O’Reilly Media, Inc., 2009. find
  38. F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, P. Prettenhofer, R. Weiss, V. Dubourg, Scikit-learn: Machine learning in Python, J. Mach. Learn. Res. 12 (2011) 2825–2830. find
  39. P. Virtanen, R. Gommers, T.E. Oliphant, M. Haberland, T. Reddy, D. Cournapeau, E. Burovski, P. Peterson, W. Weckesser, J. Bright, S.J. van der Walt, M. Brett, J. Wilson, K.J. Millman, N. Mayorov, A.R.J. Nelson, E. Jones, R. Kern, E. Larson, C.J. Carey, I. Polat, Y. Feng, E.W. Moore, J. VanderPlas, D. Laxalde, J. Perktold, R. Cimrman, I. Henriksen, E.A. Quintero, C.R. Harris, A.M. Archibald, A.H. Ribeiro, F. Pedregosa, P. van Mulbregt, SciPy 1.0: fundamental algorithms for scientific computing in Python, Nat. Methods. 17 (2020) 261–272. https://doi.org/10.1038/s41592-019-0686-2. ↗ our card
  40. X. Luo, Y. Qiao, C. Li, J. Ma, Y. Liu, An overview of microblog user geolocation methods, Inf. https://doi.org/https://doi.org/10.1016/j.ipm.2020.102375. ↗
  41. J.D.G. Paule, Y. Sun, Y. Moshfeghi, On fine-grained geolocalisation of tweets and real-time traffic incident detection, Inf. Process. Manag. 56 (2019) 1119– 1132. https://doi.org/https://doi.org/10.1016/j.ipm.2018.03.011. ↗
  42. O. Ozdikis, H. Ramampiaro, K. Nørvåg, Locality-adapted kernel densities of term co-occurrences for location prediction of tweets, Inf. Process. Manag. 56 (2019) 1280–1299. https://doi.org/https://doi.org/10.1016/j.ipm.2019.02.013. ↗
  43. P. Zola, C. Ragno, P. Cortez, A Google Trends spatial clustering approach for a worldwide Twitter user geolocation, Inf. Process. Manag. 57 (2020) 102312. https://doi.org/https://doi.org/10.1016/j.ipm.2020.102312. latitude and longitude, (2020). Process. Manag. (2020) 102375. ↗

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/alsudais-2021-similarities-between-arabic-dialects-investigating/