Elamur Research⌕
☰
home / papers / 2016 / Question Analysis for Arabic Question Answering Systems

Question Analysis for Arabic Question Answering Systems

Paper · 2016 · Waheeb Ahmed and Dr. Anto P Babu

Summary

The paper describes the question-processing front end of an Arabic question answering pipeline. It also covers how that front end feeds retrieval and answer extraction. Arabic is hard here because words carry many clitics and affixes and are usually written without short vowels, so one surface form often has several readings.

The analysis chain works in these steps:

  • It splits the question into tokens and drops prepositions, conjunctions and interrogatives.
  • It adds Arabic WordNet synonyms for nouns and adjectives.
  • It labels the question with an SVM using the two-level Li & Roth taxonomy.
  • It extracts the question focus with hand-written NP chunking rules over Stanford Arabic POS tags and parses.

Downstream, the expanded term list goes to a Vector Space Model retriever. The class label then chooses the extraction technique, and the focus is used to rank candidate answers.

On 250 questions spanning five types, the system reaches about 65% average accuracy, measured with MRR. The work matters as a compact, readable blueprint showing how much a QA system depends on its question analysis.

Key points

  • Tools: Stanford POS Tagger and Stanford Parser for Arabic, a named entity recognizer, stop-word removal, and Arabic WordNet for query expansion.
  • Arabic ambiguity: the authors note that an Arabic Tree Bank (ATB) word form averages about 2 morphological analyses. Words follow a proclitics + inflected word + enclitics shape, which makes tokenization comparable to Chinese word segmentation.
  • Question classes: the SVM classifier comes from the authors' earlier work and assigns coarse:fine labels such as DESCRIPTION:reason or NUMBER:count. The five coarse classes are HUMAN, LOCATION, NUMERIC, DESCRIPTION and ENTITY.
  • Retrieval and extraction: the retriever pulls the top 10 documents and passes 3 to answer extraction. Answers are then picked from the top 5 candidate sentences.
  • Evaluation: 250 questions of five types, about 65% average accuracy (MRR).

Interesting ideas

  • I1. The question class acts as a router: person and place questions go to NER, and date questions go to pattern matching, so each answer kind gets its own specialized extractor.
  • I2. The focus is split into a head noun plus typed modifiers (adjectival, adverbial, complement), so candidates can be ranked against structure rather than loose keywords.
  • I3. A two-level taxonomy (coarse class, then fine subtype) lets the system fall back to the general type when the fine label is uncertain.
  • I4. Synonym expansion is limited to nouns and adjectives using WordNet, which keeps verbs and function words from adding noise.
  • I5. Question analysis is judged by its effect on end-to-end answer accuracy (MRR), not only by its own intrinsic scores.

For Mova

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

  • I1 (answer): the readers could choose a dedicated answer-finder (entity lookup, quantity reasoning, time ordering) from a typed question label, and record that choice as part of the explanation.
  • I2 (frames): a question could be represented as a frame with a gap whose head and modifiers come straight from the UD subtree around the wh-word, giving a precise target to match in the text's frames.
  • I3 (answer): a coarse-then-fine question type would let a student answer with a safe broad category when the fine type is unclear, and say so explicitly.
  • I4 (world): when linking question entities to entities in a text, WordNet synonym or hypernym matches could be applied only to nominal and adjectival slots, with each match logged as an explainable link.
  • I5 (learn): parser or frame improvements could be scored by how they change fable/QA test results, which shows whether a component change actually matters.

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 5The 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 worksInformation Systems · 1,730 works

🧭 Topics 2Research subjects and languages the work is about.Semitic languages · 3,066 worksQuestion answering · 487 works

Cite

Cite as Waheeb Ahmed and Dr. Anto P Babu. 2016. Question Analysis for Arabic Question Answering Systems. arXiv:1701.02925.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.5121/ijnlc.2016.5603 · arXiv 1701.02925

@misc{ahmed2016question,
  title = {{Question Analysis for Arabic Question Answering Systems}},
  author = {Waheeb Ahmed and Dr. Anto P Babu},
  year = {2016},
  doi = {10.5121/ijnlc.2016.5603},
  eprint = {1701.02925},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.5121/ijnlc.2016.5603}
}

References

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

  1. Santosh K. and Khaled S. , “A Review and Future Perspectives of Arabic Question Answering Systems”, IEEE Transactions on Knowledge and Data Engineering, Vol. 28, pp. 3169-3190 , 2016. find
  2. Ezzeldin A.M and Shaheen M. ,” A Survey of Arabic Question Answering: challenges, Tasks, Approaches, Tools, and Future Trends”, In Proceedings of the 13th International Arab Conference on Information Technology (ACIT’2012), pp. 280-287, 2012. find
  3. J. Jonathan & K. Chunyu ,” Tokenization as The initial phase in NLP”, In Proceedings of COLING-92, NANTES, PP. 23-28, 1992.
  4. D. Jurafsky and H. James Martin, “ Speech and Language Processing: An Introduction to Natural Language Processing,” Computational Linguistics, and Speech Recognition. Prentice Hall, 2000. find
  5. Ahmed H. Aliwy,” Tokenization as Pre-processing for Arabic Tagging System”, International Journal of Information and Education Technology, Vol. 2, No. 4, pp. 348-353, 2012. find
  6. S. Bird, E. Klein, and E. Loper ,” Natural Language Processing with Python.”, O’Reilly Media, 2009. find
  7. J-P Chanod and P. Tapanainen, “A non-deterministic tokeniser for finite-state parsing,” ECAI 96. 12th European Conference on Artificial Intelligence, 1996. find
  8. N. Habash and O. Rambow, “Arabic tokenization, part-of-speech tagging and morphological disambiguation in one fell swoop,” Proceedings of the 43rd Annual Meeting of the ACL., pp. 573– 580, 2005.
  9. K. Bouzoubaa, H. Baidouri, T. Loukili, T. El Yazidi, “Arabic Stop Words:Towards a Generalisation and Standardisation”, In Proceedings of the 13th International Business Information Management Association Conference (IBIMA), 2009. find
  10. R. Al-Shalabi, G. Kanan, J.M. Jaam, and Eyad Hailat,” Stop-word removal algorithm for Arabic language”, Proceedings. Of the International Conference on Information and Communication Technologies: From Theory to Applications, 2004. find
  11. A. Mahgoub, M. Rashwan, H. Raafat, Mohamed A. Zahran, and Magda B. Fayek ”Semantic Query Expansion for Arabic Information Retrieval”, Proceedings of the EMNLP 2014 Workshop on Arabic Natural Language Processing (ANLP) , pp. 87–92, 2014.
  12. F. Noha, M. Hamdy and E. Ashraf,” Enhanced semantic arabic Question Answering system based on Khoja stemmer and AWN”, 9th International Computer Engineering Conference (ICENCO), 2013. find
  13. Hani A., Santosh K. and Khaled S., “Semantic Based Query Expansion for Arabic Question Answering Systems”, In Proceedings of the First International Conference on Arabic Computational Linguistics (ACLing), pp. 127-132, April 2015. find
  14. Global WordNet website:http://globalwordnet.org/ Arabic Wordnet-Last visited-December 2016. ↗
  15. Waheeb A. and Babu A., ”Classification of Arabic Questions Using Multinomial Naïve Bayes And Support Vector Machines”, International Journal of Latest Trends In Engineering And Technology, pp. 82-86, SACAIM, November 2016. 29 find
  16. Li X. and Roth D., “Learning Question Classifiers: The Role of Semantic Information”, Journal of Natural Language Engineering, Vol. 12, pp. 229 - 249 , September 2006. ↗
  17. http://nlp.stanford.edu/software/tagger.shtml-last visited ,December, 2016. ↗
  18. http://nlp.stanford.edu/software/lex-parser.shtml-last visited ,December, 2016. ↗
  19. Salton G., “Automatic Text Processing: the transformation, analysis, and retrieval of information by computer”, Addison Wesley, New York, ISBN:0-201-12227-8 , 1989. find
  20. R. Dukhi and A. Bhattacharya, “A Survey on Approaches to Text Mining using Information Extraction”, IOSR Journal of Computer Engineering (IOSR-JCE), pp. 27-31, 2016. find
  21. K. Shaalan and H. Raza,” NERA: Named Entity Recognition for Arabic”, Journal of The American Society for Information Science & Technology, Vol. 60, No. 8, pp. 1652–1663, 2009. find
  22. D. Moll´a and Menno V. and D. Smith, “Named Entity Recognition for Question Answering”, Proceedings of the 2006 Australasian Language Technology Workshop, pp. 57-65, 2007. find
  23. E. Noguera, A. Toral, F. Llopis, and R. Mu’noz,” Reducing question answering input data using named entity recognition”. In Proceedings of the 8th International Conference on Text, Speech & Dialogue, pages 428–434, . 2005. find
  24. A. Arara, A. Smeda, and I. Ellabib, “Searching and Analysing Arabic Text Using Regular Expressions e–Quran Case Study ”, International Journal of Computer Science and Electronics Engineering (IJCSEE), Vol. 1, pp. 627-631, 2013. find
  25. M. De Boni and S. Manandhar,” The Use of Sentence Similarity as a Semantic Relevance Metric for Question Answering”, The York Research Database, 2003. find
  26. A, Waheeb and A. Babu, “Answer Extraction and Passage Retrieval for Question Answering Systems”, International Journal of Advanced Research in Computer Engineering & Technology (IJARCET), Vol. 5, pp. 2703-2706 , 2016. find
  27. A. Waheeb and A. Babu ,” Answer Extraction Technique for Question Answering Systems”, International Journal of Innovative Research in Computer and Communication Engineering(IJIRCCE), Vol. 4, pp. 20352-20357, 2016. find
  28. E. Voorhees,” Overview of the TREC 2001 question answering track,” Proceedings of the 10th Text Retrieval Conference, pp. 42–52, 2001. Authors Waheeb Ahmed received his B.E. in Computer Science & Engineering from University of Aden, Yemen, 2008. He received his master’s degree from Osmania University, Hyderabad, 2013, under the ICCR-CE Programme. He is currently pursuing his Ph.D. in IT from Kannur University under the ICCR-GSS Programme. Dr. Babu Anto P received his Ph.D. in Technology from Cochin University of Science & Technology, India, 1992. He is currently an associate professor and working as a research guide at the Department of Information Technology, Kannur University, Kerala, India. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/ahmed-2017-question-analysis-arabic-question-answering/