Elamur Research⌕
☰
home / papers / 2013 / Development of Yes/No Arabic Question Answering System

Development of Yes/No Arabic Question Answering System

Paper · 2013 · Wafa N. Bdour and Natheer K. Gharaibeh

Summary

The paper addresses open-domain yes/no question answering for Arabic, a question type that earlier Arabic QA work (ArabiQA, QArabPro and others) had mostly left aside in favour of factoid and WH-questions. The authors build a lightweight pipeline with three modules: question analysis, text retrieval and answer selection.

Question analysis strips the interrogative particle, stop words and negation particles, recording negation as a flag. It then tags words for verb or noun and root, and uses simple rules with bottom-up parsing to separate nominal sentences (topic and comment) from verbal ones (verb and subject). The question becomes one of 12 predicate-style logical forms. In these forms, synonyms keep the polarity of the question and antonyms reverse it; the synonym and antonym lists come from the Microsoft Word Arabic thesaurus. The forms are matched against retrieved passages of variable length.

The best configuration, combining paragraph retrieval, the thesaurus and advanced search, reaches 88% correct answers with 20 documents. The work shows that a small, explicit semantic matcher can handle polar questions and serve as a starting point for richer QA.

Key points

  • The architecture has three modules (Question Analysis, Text Retrieval, Answer Selection), and Alef normalization is part of tokenization.
  • There are 12 logical representations, split by nominal vs verbal sentence and by affirmative vs negated question:
    • In nominal forms (N), the topic is kept exact and the comment is reduced to its root.
    • In verbal forms (V), the subject is kept exact and the verb is reduced to its root.
  • Query expansion uses synonyms and antonyms from the Microsoft Word Arabic thesaurus, and the article "al" is removed except in words listed in an ALEF_LAM exception file.
  • The top result is 88% correct answers with 20 documents, and paragraph retrieval beats whole-document retrieval by about 3%, especially when many documents are used.
  • About 12% error comes from the automatic tagger, which gives inconsistent roots (e.g. for words with an attached preposition) and makes noun/verb mistakes. Further errors come from the assumption that question words must appear in the answer text.

Interesting ideas

  • I1. Negation is removed from the question surface and stored as a single polarity flag on the logical form, so negated and affirmative questions share one matching path.
  • I2. Antonym expansion flips the predicate's polarity while synonym expansion keeps it, so finding "not-X" text can confirm a "Y?" question.
  • I3. Head words (topic or subject) are matched exactly, while predicates and other words are matched only by root. This is an asymmetric strictness that protects entity identity but tolerates inflection.
  • I4. Retrieval is done over variable-length paragraphs rather than whole documents, which raised precision and cut processing cost.
  • I5. The error analysis attributes most failures to a specific upstream component (root inconsistency in the tagger).

For Mova

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

  • I1 (answer): Yes/no readers on fables could compare a polarity-free frame against the text world and apply negation only at the final verdict, which keeps the explanation clean.
  • I2 (world): Antonym links from WordNet could serve as polarity-inverting edges when checking states, letting the world model answer "is the fox happy?" from a sentence saying it is sad.
  • I3 (frames): Frame matching could require exact or coreferent agent identity while allowing lemma- or class-level matches for the verb, which reduces false confirmations.
  • I4 (answer): Restricting the world-building step to retrieved paragraphs before deep analysis could speed up QA over long tales with little loss in accuracy.
  • I5 (learn): Evaluations could tag each wrong answer with the pipeline stage that caused it, which would guide where repair rules or extra silver data are most needed.

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

🎓Al-Balqa Applied UniversitySalt · Jordan2 works · 37 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksInformation Systems · 1,437 works

🧭 Topics 2Research subjects and languages the work is about.Semitic languages · 2,908 worksQuestion answering · 441 works

Cite

Cite as Wafa N. Bdour and Natheer K. Gharaibeh. 2013. Development of Yes/No Arabic Question Answering System. arXiv:1302.5675 · International Journal of Artificial Intelligence & Applications (IJAIA), Vol.4, No.1, January 2013.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.1302.5675 · arXiv 1302.5675

@misc{bdour2013development,
  title = {{Development of Yes/No Arabic Question Answering System}},
  author = {Wafa N. Bdour and Natheer K. Gharaibeh},
  year = {2013},
  doi = {10.48550/arxiv.1302.5675},
  eprint = {1302.5675},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1302.5675}
}

References

19 references, 0 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. R. Baeza-yates & B. Ribeiro-Neto, (2011) “Modern Information Retrieval”, second edition. find
  2. Breck, E., Burger, J., House, D., Light, M. & Mani, I. (1999) “Question answering from large document collections’’, Question Answering Systems: Papers from the 1999 AAAI Fall Symposium, 5-7 November, North Falmouth, MA, AAAI Press, Menlo Park, CA, pp. 26-31. find
  3. Oleksandr Kolomiyets, Marie-Francine Moens. (2011). “A survey on question answering technology from an information retrieval perspective”. Inf. Sci. 181(24): 5412-5434 . ↗
  4. Marco De Boni, Suresh Manandhar, ( 2003) “The Use of Sentence Similarity as a Semantic Relevance Metric for Question Answering”. New Directions in Question Answering , 138-144. find
  5. Monz, C. (2003). From Document Retrieval to Question Answering, ILLC Dissertation Series 2003-4, ILLC, Amsterdam . find
  6. Gaizauskas, R., Humphreys, K. (2000) .A Combined IR/NLP Approach to Question Answering Against Large Text Collections. In: 6th Content-Based Multimedia Information Access Conference (RIAO-2000), pp. 1288–1304. Paris, France. find
  7. A. Ittycheriah, M. Franz, W. Zhu, A. Ratnaparkhi, and R.J. Mammone, (2000) “IBM's Statistical Question Answering System”. In Proceedings of TREC. find
  8. Goldensohn S. , Mckeown K. , & Schlaikjer A, ( 2004) ”Answering Definitional Questions: A Hybrid Approach”, New Directions in Question Answering , 47-58. find
  9. Bailey, S. M. (2001)”Providing Question Coverage through Answer Type Classification in Natural Language Question Answering (NLQA) Systems”, Master Thesis, Georgetown University, USA, April 2001. find
  10. Huettner, A. (2000) “Question Answering”. In: 5th Search Engine Meeting. find
  11. Joho, H. (1999) ”Automatic detection of descriptive phrases for Question Answering Systems: A simple pattern matching approach”. MSc Dissertation. Department of Information Studies, University of Sheffield. Sheffield, UK. find
  12. Anne Diekema, Ozgur Yilmazel, Jiangping Chen, Sarah Harwell, Lan He, Elizabeth D. Liddy,(2003)“What do You Mean? Finding Answers to Complex Questions”, New Directions in Question Answering , 87-93. find
  13. Benajiba Y., Rosso P., Lyhyaoui A.(2007) “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 3-5. find
  14. Abouenour L., Bouzoubaa K., Rosso P. (2009). “Three-level approach for Passage Retrieval in Arabic Question /Answering Systems”. In Proc Of the 3rd International Conference on Arabic Language Processing CITALA2009, Rabat, Morocco, May, 2009. find
  15. Kanaan, G., Hammouri, A., Al-Shalabi, R. and Swalha, M.(2009). “A New Question Answering System for the Arabic Language”. American Journal of Applied Sciences 6 (4): 797-805 ↗
  16. Akour M., Abufardeh S., Magel K. and Al-Radaideh Qasem A., (2011),”QArabPro: A Rule Based Question Answering System for Reading Comprehension Tests in Arabic”, American Journal of Applied Sciences, 8(6): pp. 652-661. find
  17. Ditters, Everhard.(2001)”A Formal Grammar for the Description of Sentence Structure in Modern Standard Arabic”, In EACL 2001 Workshop Proceedings on Arabic Language Processing: Status and Prospects, pages 31–37, Toulouse, France. find
  18. Alllen J. (1994),”Natural language understanding” , Benjamin/Cummings, 2nd edition find
  19. HADDAD Bassam & YASEEN Mustafa (2001), “Towards Understanding Arabic: A Logical Approach for Semantic Representation”, in Dichy, Krauwer & Yaseen, eds., 2001: 44-50. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/bdour-2013-development-yes-no-arabic-question/