Elamur Research⌕
☰
home / papers / 2016 / Extracting Arabic Relations from the Web

Extracting Arabic Relations from the Web

Paper · 2016 · Shimaa M. Abd El-salam, Enas M.F. El Houby, A.K. Al Sammak et al.

Summary

Arabic text on the Web holds many facts about named entities, but they sit in loosely organised prose and cannot be queried directly. Annotated Arabic corpora for relation extraction are scarce. The authors present what they describe as the first semi-supervised (bootstrapping) system that harvests binary relations between two Arabic named entities. The first entity is always a person, and the second can be any other entity type. A user supplies a few seed pairs. The system sends them to Google and treats the returned result snippets as its corpus, learns surface patterns around the seed pairs, and applies those patterns to find new pairs. New pairs with high credibility scores become seeds for the next round. The loop stops once a user-chosen number of instances has been collected. Entities are recognised from the learned patterns themselves, so no entity gazetteer is needed. Syntactic analysis of the Arabic text improves the output. Four relation types were tested. Precision fell between 0.61 and 0.75 and recall between 0.71 and 0.83, with (player, club) performing best. The approach matters because it builds domain fact tables cheaply for a language with few resources.

Key points

  • The input is a handful of seed pairs (e1, e2) for one relation, and the source text is Google search result summaries rather than full web pages.
  • Each iteration has two phases: pattern extraction, then instance extraction. Only relations whose two entities appear on the same line of text are targeted.
  • Credibility scoring decides which extracted instances are promoted to seeds. A user-defined instance count ends the iteration.
  • Across four experiments, precision was 0.61–0.75 and recall was 0.71–0.83. The (player, club) relation reached 0.72 precision and 0.83 recall.
  • The listed Arabic-specific difficulties include:
    • several relations holding between the same entity pair;
    • implicit relations such as Belong-to;
    • one relation element (NE1, R or NE2) missing from the sentence and recoverable only from earlier context.

Interesting ideas

  • I1. Use search-engine snippets as the corpus: short text windows already centred on the query terms give dense pattern-learning material without crawling whole pages
  • I2. Promote only high-credibility outputs back into the seed pool, so the bootstrap grows while limiting semantic drift
  • I3. Infer entity boundaries from the learned patterns instead of relying on a prepared name dictionary, so the slot a pattern fills defines what counts as an entity
  • I4. Treat a missing relation element (NE1, R, ?) as something to resolve from preceding context rather than discarding the sentence
  • I5. Keep the system relation-agnostic, so changing the seed pairs alone retargets it to a new domain such as director–film or singer–song

For Mova

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

  • I1 (learn): Snippet-sized windows could serve as cheap, focused contexts for harvesting silver examples of a single verb frame or relation before witness parsers vote on them.
  • I2 (learn): Self-induced repair rules could adopt the same gate, letting a newly learned rule feed later rounds only after it scores above a confidence threshold on held-out cases.
  • I3 (world): Unknown names could be typed in the world model by the frame slots they repeatedly fill, such as "plays for X" marking X as a club, with the evidence kept for explanation.
  • I4 (frames): Frames with an empty role could be flagged and completed through coreference with earlier sentences, with each filled slot recording which antecedent supplied it.
  • I5 (world): Populating the knowledge graph for a new domain could begin from a few typed seed facts that drive pattern harvesting over UD trees rather than hand-written extractors.

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

🎓Benha UniversityBanhā · Egypt14 works · 89 citations🔬National Research CentreGiza · Egypt2 works · 11 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 1Research subjects and languages the work is about.Semitic languages · 3,066 works

Cite

Cite as Shimaa M. Abd El-salam, Enas M.F. El Houby, A.K. Al Sammak and T.A. El-Shishtawy. 2016. Extracting Arabic Relations from the Web. arXiv:1603.02488.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.5121/ijcsit.2016.8107 · arXiv 1603.02488

@misc{elsalam2016extracting,
  title = {{Extracting Arabic Relations from the Web}},
  author = {Shimaa M. Abd El-salam, Enas M.F. {El Houby}, A.K. {Al Sammak} and T.A. El-Shishtawy},
  year = {2016},
  doi = {10.5121/ijcsit.2016.8107},
  eprint = {1603.02488},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.5121/ijcsit.2016.8107}
}

References

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

  1. T.Wang, Yaoyong Li, K. Bontcheva, H. Cunningham and J. Wang, (2006) “Automatic Extraction of Hierarchical Relations from Text”. In Springer-Verlag Berlin Heidelberg IJCNLP 2006, LNcs 4011, pp. 215 –229. find
  2. Grishman, R., (1997) “Information Extraction: Techniques and Challenges”. In Pazienza, M.T. (ed.) SCIE 1997. LNCS, vol. 1299, pp. 10–27. Springer, Heidelberg. find
  3. Bach N., and Badaskar S., (2007) “A review of relation extraction”. In Literature review for Language and Statistics II. find
  4. Abreu, S.C., Bonamigo, T.L. and Vieira, R., (2013) “A review on relation extraction with an eye on Portuguese”. In Journal of the Brazilian Computer Society 19, 553–571, 12 Jun 2013. Available online 9 July. ↗
  5. Chao Chen, Liang He and Xin Lin, (2012) “REV: Extracting Entity Relations from World Wide Web”. In proceeding of ACM, 978-1-4503-1172-4. ICUIMC’12, February 20–22, Kuala Lumpur, Malaysia. find
  6. Xiaoling Yang, Jing Yang, and Chao Chen.L., (2012) “Tuple Refinement Method Based on Relationship Keyword Extension”. In WISM Springer-Verlag Berlin Heidelberg, LNCS 7529, pp. 564–571. find
  7. Qingliang Miao, Shu Zhang, Bo Zhang, Yao Meng and Hao Yu, (2012) “Extracting and Visualizing Semantic Relationships from Chinese Biomedical Text”. In proceeding of ACM 26th Pacific Asia Conference on Language, Information and Computation pages 99-107. find
  8. Abdelmajid Ben Hamadou, Odile Piton and Héla Fehri, (2010) “Multilingual Extraction of functional relations between Arabic Named Entities using NooJ platform”. In Hal-00547940, version 1, 17 Dec. Université Paris I - Panthéon Sorbonne. find
  9. Belkacem K. and Badr A., (2011) “Extracting Entities and Relationships from Arabic Text for Information System”. In Journal of Emerging Trends in Computing and Information Sciences (CIS). VOL. 2, NO. 11, October 2011, ISSN 2079- 8407. find
  10. Fatma Ali Alnairia, Nazlia Omar and Mohammed Albared, (2012) “Extraction of Spatial Relation in Arabic Text Using Rule-Based Approach”. In proceeding International Journal of Advancements in Computing Technology (IJACT) Volume4, Number15, September. ↗
  11. Alotayq, A. (2013) “Extracting relations between Arabic named entities”. In proceeding: TSD2013. Springer-Verlag, Berlin Heidelberg, Pilsen, pp. 265– 271. find
  12. aInes Boujelben, Salma Jamoussi and Abdelmajid Ben Hamadou, (2014) “RelANE: Discovering Relations between Arabic Named Entities”. In Proceedings for Springer International Publishing Switzerland. (Eds.): TSD 2014, LNAI 8655, pp. 233–239. find
  13. bInes Boujelben, Salma Jamoussi and Abdelmajid Ben Hamadou, (2014) “A hybrid method for extracting relations between Arabic named entities”. In Journal of King Saud University – Computer and Information Sciences, available online 28 September 2014. 26, 425–440. ↗
  14. Elkateb, S., Black, W., Rodriguez, H., Alkhalifa, M., Vossen, P., Pease, A., and Fellbaum, C., (2006) “Building a WordNet for Arabic”. In Proceedings of the Fifth International Conference on Language Resources and Evaluation, Genoa, Italy.
  15. Santos D., Mamede N., and Baptista J., (2010) “Extraction of Family Relations between Entities”. In proceedings INForum 2010 - II Simp´osio de Inform´atica, Lu´ıs S. Barbosa, Miguel P. Correia (eds), 9-10 Setembro, pp. 549–560. find
  16. Mohammed G.H. Al Zamil and Qasem Al-Radaideh (2014) “Automatic extraction of ontological relations from Arabic text”. In Journal of King Saud University – Computer and Information Sciences. 26, 462–472 Department of Computer Information Systems, Yarmouk University, Irbed, Jordan Available online 28 September. ↗
  17. Hasegawa, T., Sekine, S. and Grishman, R. (2004) “Discovering relations among named entities from large corpora”. In Proceedings of Association for Computational Linguistics. Morris-town, NJ, USA
  18. Zhang M., Su, J., Wang, D.Zhou, G., Tan, and C.L. (2005) “Discovering Relations between Named Entities from a Large Raw Corpus Using Tree Similarity-Based Clustering”. In Proceeding of IJCNLP ’05, LNAI, vol. 3651, pp. 378–389. find
  19. Brin, S. (1998). "Extracting patterns and relations from the world wide web". In WebDB. LNCS, vol. 1590, pp. 172–183. Springer, Heidelberg Workshop at 6th International Conference on Extending Database Technology, EDBT’98. find
  20. Agichtein, E., Gravano, L., (2000) "Snowball: Extracting relations from large plain-text collections". In Proceedings of the 5th ACM International Conference on Digital Libraries. ↗
  21. Deepak Ravichandran and Eduard H. Hovy, (2002) “Learning surface text patterns for a question answering system”. In proceeding of the 40th annual meeting of the ACL- July, pp. 41–47, Philadelphia, PA.
  22. Marco Pennacchiotti and Patrick Pantel, (2006) “A bootstrapping algorithm for automatically harvesting semantic relations”. In Proceedings of Inference in Computational Semantics (ICoS-06), pages 87–96. find
  23. Benjamin Rozenfeld and Ronen Feldman, (2008) “Self-supervised relation extraction from the web”. In proceeding of Knowledge and Information Systems, 17(1):17–33. ↗
  24. Oren Etzioni, Michael Cafarella, Doug Downey, Ana-Maria Popescu Tal Shaked, Stephen Soderland, Daniel S.Weld, and Alexander Yates, (2005) “unsupervised named-entity extraction from the Web: an experimental study”. In Artif Intell 165(1):91–134. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/el-salam-2016-extracting-arabic-relations-web/