Elamur Research⌕
☰
home / papers / 2018 / Towards an Automatic Text Comprehension for the Arabic Question-Answering: Semantic and Logical Representation of Texts

Towards an Automatic Text Comprehension for the Arabic Question-Answering: Semantic and Logical Representation of Texts

Paper · 2018 · Wided Bakari, Patrice Bellot and Mahmoud Neji

Summary

The paper addresses Arabic question answering, where the goal is to return a precise answer rather than a list of documents. Arabic is hard to process: its morphology is rich, word order is fairly free, there is no capitalization, short vowels are usually omitted, and it lacks resources like WordNet. The authors propose a pipeline that tries to "understand" both the question and candidate passages. Each text becomes a Sowa-style conceptual graph built from terms, concepts and the relations between them. An algorithm based on Sowa's Φ operator then turns each graph into a first-order logic formula. Answering is framed as recognizing textual entailment: the passage plays the role of the text and the question the role of the hypothesis. Three kinds of features feed a WEKA J48 decision tree, which makes a yes/no decision. Passages judged entailing are ranked by a confidence score. On 250 factual questions with 2500 web passages, the system reaches 74% accuracy. This shows that structured semantic and logical representations are workable even for a resource-poor language.

Key points

  • Each question and passage is mapped to a conceptual graph, a finite, connected, bipartite graph with concept nodes and relation nodes. The graph is then converted to first-order logic via Sowa's (1984) Φ operator.
  • In the logical form, concept types become unary predicates and relation types become predicates of matching arity. Individual markers (e.g. ALI) become constants, generic markers (*) become variables, and everything is joined by conjunction.
  • Entailment uses three features: overlap of predicate-arguments, matching of named entities, and semantic similarity (e.g. Wu-Palmer). A WEKA J48 tree classifies each pair as "YES" or "NO".
  • The evaluation set has 250 factual questions from forums, FAQs, TREC and CLEF, translated from English into Arabic. Its 2500 passages were retrieved through Google, and the reported accuracy is 74%.
  • The related-work survey covers Arabic QA systems from AQAS (1993) to Lemaza (2017).

Interesting ideas

  • I1. QA is recast as entailment: the passage is the premise, the question is the hypothesis, and only passages that pass a yes/no entailment check become answer candidates. This splits answering into a clean verification step plus a ranking step.
  • I2. Graphs are flattened to logic by a fixed recipe: one existentially bound variable plus a unary predicate per node, one binary predicate per edge, and a conjunction over all of them. Because the recipe is deterministic, it is trivially explainable and reversible.
  • I3. Named individuals map to constants and unspecified referents ("some person") map to variables, so the logic keeps the difference between definite and generic mentions. This matters when deciding whether two texts talk about the same entity.
  • I4. Entailment rests on a handful of interpretable feature scores, namely argument overlap, entity match and taxonomy-based similarity, fed to a small decision tree instead of an opaque model. Each decision can therefore be traced to concrete feature thresholds.
  • I5. Wu-Palmer similarity over a lexical taxonomy is used as a soft matching signal between predicates, so near-synonyms can still align.

For Mova

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

  • I1 (answer): A fable or school-test reader could first check whether a sentence in the world of the text supports the question, and only then extract the answer span, logging the verdict as part of the explanation.
  • I2 (world): Frame graphs could be serialized into conjunctive predicate form this way, giving a canonical, diffable representation of the world of the text for matching and debugging.
  • I3 (frames): When frames are built, coreference-resolved entities could become constants and unresolved or generic arguments could become variables, so later matching knows which slots may unify freely.
  • I4 (learn): A J48-style tree over a few structural match scores could serve as an inspectable baseline or negative control against which heavier learned answer-checkers are compared.
  • I5 (world): WordNet-based Wu-Palmer scores could relax predicate matching between question and text frames, while the exact similarity value is kept in the explanation trace.

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 SfaxSfax · Tunisia17 works · 233 citations🔬Laboratoire d’Informatique et SystèmesMarseille · France10 works · 187 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 worksQuestion answering · 487 works

Cite

Cite as Wided Bakari, Patrice Bellot and Mahmoud Neji. 2018. Towards an Automatic Text Comprehension for the Arabic Question-Answering: Semantic and Logical Representation of Texts. Proceedings of the 32nd Pacific Asia Conference on Language, Information and Computation.

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

@inproceedings{bakari2018automatic,
  title = {{Towards an Automatic Text Comprehension for the Arabic Question-Answering: Semantic and Logical Representation of Texts}},
  author = {Wided Bakari, Patrice Bellot and Mahmoud Neji},
  booktitle = {Proceedings of the 32nd Pacific Asia Conference on Language, Information and Computation},
  year = {2018},
  url = {https://aclanthology.org/Y18-1006/}
}

References

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

  1. Banerjee, S., Bhaskar, P., Pakray, P., Bandyopadhyay, S., & Gelbukh, A. F. (2013, September). Multiple Choice Question (MCQ) Answering System for Entrance Examination. In CLEF (Working Notes). find
  2. Vanderwende, L. (2007, March). Answering and Questioning for Machine Reading. In AAAI Spring Symposium: Machine Reading (p. 91). find
  3. Patil, N. V., Patil, A. S., & Pawar, B. V. Issues and Challenges in Marathi Named Entity Recognition. International Journal on Natural Language Computing (IJNLC) Vol, 5, 15-30. 2016. find
  4. Gómez-Adorno, H., Pinto, D., & Ayala, D. V. (2013). Semantic Answer Validation in Question Answering Systems for Reading Comprehension Tests. In AMW. find
  5. Sowa John F. (1984). Conceptual Structures: Information Processing in Mind and Machine. Addison-Wesley Company. find
  6. Clark, P., Harrison, P. Recognizing Textual Entailment with Logical Inference. In Proceedings of 2008 Text Analysis Conference (TAC’08), Gaithsburg, Maryland, 2008. find
  7. Mohammed, F. A., Nasser, Khaled, et Harb, H. M. A knowledge based Arabic question answering system (AQAS). ACM SIGART Bulletin, 1993, vol. 4, no 4, p. 21-30. find
  8. Hammo, B., Abuleil, S., Lytinen, S., & Evens, M. (2004). Experimenting with a question answering system for the Arabic language. Computers and the Humanities, 38(4), 397-415. ↗
  9. Benajiba, Y., Rosso, P., & Lyhyaoui, A. (2007, April). Implementation of the ArabiQA question answering system’s components. In Proc. Workshop on Arabic Natural Language Processing, 2nd Information Communication Technologies Int. Symposium, ICTIS-2007, Fez, Morroco, April (pp. 3-5). find
  10. Brini, W., Ellouze, M., & Hadrich Belguith, L. (2009). QASAL: Un système de question-réponse dédié pour les questions factuelles en langue Arabe. 9ème Journées Scientifiques des Jeunes Chercheurs en Génie Electrique et Informatique, Tunisia. find
  11. Trigui, O., Belguith, L. H., & Rosso, P. (2010). DefArabicQA: Arabic definition question answering system. In Workshop on Language Resources and Human Language Technologies for Semitic Languages, 7th LREC, Valletta, Malta (pp. 40-45). find
  12. Kanaan, G., Hammouri, A., Al-Shalabi, R. and Swalha, M. “A New Question Answering System for the Arabic Language”, American Journal of Applied Sciences 6 (4). 2009. 797-805. ↗
  13. Abouenour, L., Bouzoubaa, K., & Rosso, P. (2012). IDRAAQ: New Arabic Question Answering System Based on Query Expansion and Passage Retrieval. In CLEF (Online Working Notes/Labs/Workshop). find
  14. Abouenour, L. (2014). Three-levels Approach for Arabic Question Answering Systems. Diss. Ecole Mohammadia d’Ingénieurs. find
  15. Bdour W. N and Gharaibeh N. K. (2013) ―Development of Yes/No Arabic Question Answering System‖, In International Journal of Artificial Intelligence & Applications (IJAIA),. Vol.4. No.1, January 2013. DOI: 10.5121/ijaia.2013.4105 51. ↗
  16. Ezzeldin, A. M., Kholief, M. H., & El-Sonbaty, Y. (2013, September). ALQASIM: Arabic language question answer selection in machines. In International Conference of the Cross-Language Evaluation Forum for European Languages (pp. 100-103). Springer, Berlin, Heidelberg. find
  17. Bekhti, S., Rehman, A., AL-Harbi, M. and Saba, T. “AQUASYS: an arabic question-answering system based on extensive question analysis and answer 50 32nd Pacific Asia Conference on Language, Information and Computation Hong Kong, 1-3 December 2018 Copyright 2018 by the authors relevance scoring”, In International Journal of Academic Research; Jul2011, Vol. 3 Issue 4, p45. and the Arabic language, November 2007, Riyadh, Saudi Arabia. find
  18. Kurdi, H., Alkhaider, S., & Alfaif, N. (2014). Development and evaluation of a web based question answering system for arabic language. Computer Science & Information Technology (CS & IT), 4(02), 187-202. find
  19. Almarwani, Nada and Diab, Mona. Arabic Textual Entailment with Word Embeddings. In : Proceedings of the Third Arabic Natural Language Processing Workshop. 2017. p. 185-190.
  20. Abdelnasser, H., Ragab, M., Mohamed, R., Mohamed, A., Farouk, B., El-Makky, N., & Torki, M. (2014). Al-Bayan: an arabic question answering system for the holy quran. In Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing (ANLP)(pp. 57-64).
  21. Chklovski, Timothy and Pantel, Patrick. Verbocean: Mining the web for fine-grained semantic verb relations. In : Proceedings of the 2004 Conference on Empirical Methods in Natural Language Processing. 2004. find
  22. Bakari, W., Bellot, P., & Neji, M. (2016). AQA-WebCorp: Web-based factual questions for Arabic. Procedia Computer Science, 96, 275-284. ISO 690 find
  23. Ian H. Witten and Eibe Frank. 1999. Data Mining: Practical Machine Learning Tools and Techniques with Java Implementations. Morgan Kaufmann. find
  24. Al-Khawaldeh, F. "Answer Extraction for Why Arabic Question Answering Systems: EWAQ," Proc. World of Computer Science and Information Technology Journal (WCSIT), vol. 5, no. 5, pp. 82-86, 2015. find
  25. Al-Shawakfa, E. (2016). A Rule-Based Approach to Understand Questions in Arabic Question Answering. Jordanian J. Comput. Inf. Technol, 2(3), 210-231. ↗
  26. Azmi, A. M., & Alshenaifi, N. A. (2017). Lemaza: An Arabic why-question answering system. Natural Language Engineering, 23(6), 877-903. ↗
  27. Rosso, P., Benajiba, Y. and Lyhyaoui, A. “Towards an Arabic Question Answering system”, In Proceedings of 4th Conference on Scientific Research Outlook & Technology Development in the Arab world. SROIV. Damascus, Syria. 11-14 December. 2006. find
  28. Hamdelsayed, M., & Atwell, E. S. (2016). Islamic Applications of Automatic Question-Answering. Journal of Engineering and Computer Science, 17(2), 51-57. find
  29. Miller, George A. WordNet: a lexical database for English. Communications of the ACM, 1995, vol. 38, no 11, p. 39-41.
  30. AlAgha I, Abu-Taha A (2015) AR2SPARQL: An Arabic Natural Language Interface for the Semantic Web, International Journal of Computer Applications, 125(6), pp 19-27. find
  31. Al-Khalifa H, Al-Wabil A (2007) The Arabic language and the semantic web: Challenges And opportunities, In: The first international symposium on computers 51 32nd Pacific Asia Conference on Language, Information and Computation Hong Kong, 1-3 December 2018 Copyright 2018 by the authors find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/bakari-2018-automatic-text-comprehension-arabic-question/