Elamur Research⌕
☰
home / papers / 2014 / Evaluation of a possibilistic classification approach for Arabic texts disambiguation (Evaluation d'une approche de classification possibiliste pour la désambiguïsation des textes arabes) [in French]

Evaluation of a possibilistic classification approach for Arabic texts disambiguation (Evaluation d'une approche de classification possibiliste pour la désambiguïsation des textes arabes) [in French]

Paper · 2014 · Raja Ayed, Ibrahim Bounhas, Bilel Elayeb et al.

Summary

Unvocalized Arabic is highly ambiguous at the morphological level. Without short vowels, a single written form can admit more than a dozen analyses. For example, wqf can be read as "he stood", "endowment", or "and stand!". The authors treat morphological disambiguation as classification. The task is to choose the correct value of each morphological attribute from the candidates a morphological analyzer proposes. Whereas most prior taggers address POS alone, this work handles every one of the 14 attributes laid out previously by Habash and colleagues (2009) and by Ayed and colleagues (2012b).

They propose three models built on possibilistic classification, which uses possibility and necessity measures from possibility theory. These models tolerate imprecise data in both training and testing, because the training material is untagged text rather than gold annotation. Morphological dependencies are learned from vocalized texts and then applied to unvocalized ones. Experiments use the Hadith corpus (classical Arabic) and the Arabic Treebank (modern Arabic). The models are compared with probabilistic and statistical classifiers, and the training and test sets are restructured so those baselines can handle imprecise instances. The excerpt provided does not include the numeric results.

Key points

  • The task is morphological disambiguation of unvocalized Arabic. It covers 14 morphological attributes (POS, gender, number, etc.), not just POS.
  • Three possibilistic classification models are proposed. They are trained on untagged, vocalized text and tested on unvocalized text.
  • The models rely on possibility theory: a distribution π over states in [0,1], the possibility measure Π(A) = max π(ω), and a derived necessity measure.
  • Evaluation uses two corpora of different registers: the Hadith corpus (classical) and the Arabic Treebank (modern).
  • The baselines are probabilistic and statistical classifiers. Their training and test sets are converted to cope with instances whose labels are imprecise.

Interesting ideas

  • I1. When an instance can only be pinned to a set of candidate labels, a possibility distribution represents that honestly instead of forcing a single gold tag. Possibility and necessity then give separate scores for "not excluded" and "strongly supported."
  • I2. Text that carries extra explicit marking (vocalized Arabic) can serve as a free, partially disambiguated training signal. A model learned there can be applied to the underspecified everyday form.
  • I3. Disambiguation is framed as choosing among an analyzer's candidate set for each attribute separately, rather than open-ended tagging. This keeps every decision traceable to an explicit list of alternatives.
  • I4. Testing on both a classical and a modern corpus shows whether learned dependencies carry over across registers.
  • I5. To compare fairly with standard classifiers, the authors converted imprecise multi-candidate instances into a form those classifiers can consume.

For Mova

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

  • I1 (tokens): The tagger could keep a possibility/necessity pair per candidate lemma or feature. Downstream stages could then see which choices are firmly supported and which are merely not ruled out.
  • I2 (learn): Silver data could come from texts where ambiguity is already reduced by extra markup or punctuation. Agreement could then be measured on plain text, with each instance keeping its residual candidate set.
  • I3 (tokens): Morphology could output a candidate set per feature (number, tense, person), and the tagger would choose within it. That makes each selection explainable by which candidates were eliminated.
  • I4 (learn): Evaluation could routinely pair a fable/tale set with a modern news-style set to reveal register-specific overfitting in the tagger or parser.
  • I5 (learn): When benchmarking a set-valued component against single-label baselines, the conversion step should be stated explicitly so negative controls and baselines see the same information.

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, topics and concepts

📚 Shelves 2The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 works

🧭 Topics 4Research subjects and languages the work is about.Semitic languages · 3,066 worksEvaluation & benchmarks · 1,341 worksMorphology · 420 worksFrench · 83 works

Cite

Cite as Raja Ayed, Ibrahim Bounhas, Bilel Elayeb, Narjès Bellamine Ben Saoud and Fabrice Evrard. 2014. Evaluation of a possibilistic classification approach for Arabic texts disambiguation (Evaluation d'une approche de classification possibiliste pour la désambiguïsation des textes arabes) [in French]. Proceedings of TALN 2014 (Volume 1: Long Papers).

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

@inproceedings{ayed2014evaluation,
  title = {{Evaluation of a possibilistic classification approach for Arabic texts disambiguation (Evaluation d'une approche de classification possibiliste pour la désambiguïsation des textes arabes) [in French]}},
  author = {Raja Ayed, Ibrahim Bounhas, Bilel Elayeb, Narjès Bellamine Ben Saoud and Fabrice Evrard},
  booktitle = {Proceedings of TALN 2014 (Volume 1: Long Papers)},
  year = {2014},
  url = {https://aclanthology.org/F14-1028/}
}

References

37 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. A YED R., B OUNHAS I., E LAYEB B., E VRARD F., B ENLLAMINE B EN S AOUD N. (2012a). A Possibilistic Approach th for the Automatic Morphological Disambiguation of Arabic Texts. In: T. Hochin & R. Lee (Eds.), Proceedings of the 13 ACIS International Conference on Software Engineering, Artificial Intelligence, Networking and Parallel Distributed Computing (SNPD), Kyoto, Japan,187-194. find
  2. A YED R., B OUNHAS I., E LAYEB B., E VRARD F., B ENLLAMINE B EN S AOUD N. (2012b). Arabic Morphological Analysis and Disambiguation Using a Possibilistic Classifier. In Intelligent Computing Theories and Applications, Proceedings of the 8 th International Conference on Intelligent Computing (ICIC), China, 274-279. find
  3. B OUNHAS I., S LIMANI Y. (2009). A hybrid approach for Arabic multi-word term extraction. In Proceedings of the International Conference on Natural Language Processing and Knowledge Engineering (NLP-KE), Dalian, China, 1–8. find
  4. B OUNHAS I., E LAYEB B., E VRARD F., S LIMANI Y. (2011). Organizing Contextual Knowledge for Arabic Text Disambiguation and Terminology Extraction. Knowledge Organization 38(6):473–490. find
  5. C OLLINS M. (2002). Discriminative training methods nd for hidden Markov models: theory and experiments with perceptron algorithms. In Proceedings of the ACL-2 conference on Empirical methods in natural language processing, Stroudsburg, PA, USA, 1-8. find
  6. D AOUD D. (2009). Synchronized Morphological and Syntactic Disambiguation for Arabic. Advances in Computational Linguistics 41, 73-86. find
  7. D IAB M., H ACIOGLU K., J URAFSKY D. (2004). Automatic Tagging of Arabic Text: From Raw Text to Base Phrase Chunks. In Human Language Technology conference / North American chapter of the Association for Computational Linguistics annual meeting, Boston, USA, 149-152.
  8. D UBOID D., P RADE H. (1985). héorie des possibilités: applications à la représentation des connaissances en informatique. Masson, Paris, France. find
  9. D UBOID D., P RADE H. (1994). Possibility Theory: An Approach to computerized Processing of Uncertainty. Plenum Press, New York, USA. find
  10. E LAYEB B., E VRARD F., Z AGHDOUD M., B EN A HMED M. (2009). Towards an intelligent possibilistic web information retrieval using multiagent system. Interactive Technology and Smart Education 6(1): 40–59. find
  11. E L H ADJ Y., A L -S UGHAYEIR I., A L -A NSARI A. (2009). Arabic Part-Of-Speech Tagging using the Sentence Structure. In: Proceedings of the Second International Conference on Arabic Language Resources and Tools, Cairo, Egypt, 241-245. find
  12. E LSHAFEI M., A L -M UHTASEB H., A L -G HAMDI M. (2002). Techniques for high quality Arabic speech synthesis. Information Sciences 140(3), 255-267. find
  13. F ETTWEIS G., M EYR H. (1991). High-speed parallel Viterbi decoding: algorithm and VLSI-architecture. IEEE Communications Magazine, 46- 55. find
  14. F ORNEY G.D. (1973). The Viterbi algorithm. Proceedings of IEEE 61: 268-278. find
  15. G ARSIDE R., L EECH F. (1987). The UCREL probabilistic parsing System. The Computational Analysis of English: A Corpus-Based Approach, Longman, London, 66-81. find
  16. H ABASH N., R AMBOW O. (2005). Arabic Tokenization, Part-of-speech Tagging and Morphological Disambiguation in One Fell Swoop. In: the proceedings of the 43 rd Annual Meeting on Association for Computational Linguistics, Stroudsburg, PA, USA, 573–580.
  17. H ABASH N., R AMBOW O. (2007). Arabic Diacritization Through Full Morphological Tagging. In: Human Language Technologies: The Conference of the North American Chapter of the Association for Computational Linguistics, Stroudsburg, PA, USA, 53-56.
  18. H ABASH N., R AMBOW O., R OTH R. (2009). Mada+tokan: A toolkit for Arabic tokenization, diacritization, morphological disambiguation, pos tagging, stemming and lemmatization. In Proceedings of the 2nd International Conference on Arabic Language Resources and Tools (MEDAR). find
  19. H AJIC J. (2000). Morphological Tagging: Data vs. Dictionaries. In: Proceedings of the 1 st North American Chapter of the Association for Computational Linguistics Conference, Stroudsburg, PA, USA, 94-101.
  20. H AOUARI B., B EN A MOR N., E LOUEDI Z., M ELLOULI K. (2009). Naïve possibilistic network classifiers. Fuzzy Sets and Systems 160(22): 3224-3238. find
  21. H ARRAG F., A LOTHAIM A., A BANMY A., A LOMAIGAN F., A LSALEHI S. (2013). Ontology Extraction Approach for Prophetic Narration (Hadith) using Association Rules. International Journal on Islamic Applications in Computer Science And Technology 1(2): 48-57. find
  22. H OCEINI Y., C HERAGUI M. A., A BBAS M. (2011). Towards a New Approach for Disambiguation in NLP by Multiple Criterian Decision-Aid. The Prague Bulletin of Mathematical Linguistics 95, 19-32. find
  23. J URAFSKY D., M ARTIN J.H. (2009). Speech and language processing: an introduction to natural language processing, computational linguistics, and speech recognition. Pearson Prentice Hall, Upper Saddle River, New Jersey, USA. find
  24. K ESKES I., B EANAMARA F., H ADRICH B ELGUITH L. (2013). Segmentation de textes arabes en unités discursives minimales. TALN-RECITAL, Les sables d’Olonne, 435-449. find
  25. K HOJA S H . (2001). APT: Arabic part-of-speech tagger. In: Proceedings of Student Workshop at the Second Meeting of the North American Association for Computational Linguistics, Carnegie Mellon University, Pennsylvania, USA. find
  26. K OHAVI R. (1995). A th Study of Cross-validation and Bootstrap for Accuracy Estimation and Model Selection. In: Proceedings of the 14 International Joint Conference on Artificial Intelligence, San Francisco, CA, USA, 1137-1143. find
  27. L IN J., V ITTER S, J., H ELLERSTEIN L . (1994). A Theory for Memory-Based Learning. Machine Learning 17(2-3): 143-167. find
  28. M AAMOURI M., B IES A., K ULICK S. (2009). Creating a Methodology for Large-Scale Correction of Treebank Annotation: The Case of the Arabic Treebank. In the proceedings of MEDAR Second International Conference on Arabic Language Resources and Tools, Cairo, Egypt, 138–144. find
  29. M ANSOUR S., S IMA ’ AN K., W INTER Y. (2007). Smoothing a lexicon-based pos tagger for Arabic and Hebrew. ACL07 Workshop on Computational Approaches to Semitic Languages, Prague, Czech, 97-103.
  30. O THMAN E., S HAALAN K., R AFEA A. (2004). Towards Resolving Ambiguity in Understanding Arabic Sentence. In the proceedings of International Conference on Arabic Language Resources and Tools, NEMLAR, Egypt, 118-122. find
  31. P EARL J. (1988). Probabilistic reasoning in intelligent systems: networks of plausible inference. Probabilistic Reasoning in Intelligent Systems. Morgan Kaufmman, San Francisco, California, USA. find
  32. Q UINLAN J. R. (1986). Induction of decision trees. Machine Learning 1(1): 81-106. find
  33. R OTH R., R AMBOW O., H ABASH N., D IAB M., R UDIN C. (2008). Arabic Morphological Tagging, Diacritization, and Lemmatization Using Lexeme Models and Feature Ranking. In: Proceedings of the Association for Computational Linguistics conference (ACL), Columbus, Ohio, USA, 117-120.
  34. S CHMID H. (1994). Probabilistic part-of-speech tagging using decision trees. In Proceedings of the International Conference on New Methods in Language Processing, Manchester, UK, 44-49. find
  35. T LILI -G UIASSA Y. (2006). Hybrid Method for Tagging Arabic Text. Journal of Computer Science 2(3): 245-248. find
  36. V APNIK V. (1998). Statistical Learning Theory. Wiley, New York, USA, 1-736. find
  37. Z RIBI C., T ORJMEN A., B EN A HMED th M. (2006). An Efficient Multi-agent System Combining POS-Taggers for Arabic Texts. In Proceedings of 7 international conference of Computational Linguistics and Intelligent Text Processing, LNCS Volume 3878, Springer, 121-131. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/ayed-2014-evaluation-possibilistic-classification-approach-arabic/