Elamur Research⌕
☰
home / papers / 2013 / A Semi-supervised Learning Approach to Arabic Named Entity Recognition

A Semi-supervised Learning Approach to Arabic Named Entity Recognition

Paper · 2013 · Maha Althobaiti, Udo Kruschwitz and Massimo Poesio

Summary

Arabic NER has mostly relied on hand-written rules or on supervised learners. Both need annotated corpora or gazetteers, which are costly to build and hard to extend to new entity categories. The authors present ASemiNER, a bootstrapping system that starts from a handful of seed names per class plus trigger nouns and verbs mined semi-automatically from Arabic Wikipedia.

Around each seed occurrence it collects a window of tokens and abstracts it into a POS-based surface pattern. Only trigger words and prepositions stay as literal tokens. Regexes built from these patterns harvest new candidate names, which are ranked, and the best are fed back as seeds. Training uses ANERcorp, later combined with ACE 2005, and evaluation uses the ANERcorp test set.

The tuned system beats a maximum-entropy supervised baseline but not stronger supervised systems. It also moves with little effort to finer person subtypes. This matters because it shows workable NER with minimal supervision in a language without capitalisation cues.

Key points

  • Best configuration: 20 seeds per class and 10 iterations, with candidates ranked by how many distinct patterns extracted them; going beyond 10 iterations brought no notable gain.
  • Context windows of 7 tokens were used, a value chosen after trials; verbs are reduced to roots, while nouns keep their inflected form.
  • PMI was tried as an alternative ranking measure and performed worse than the distinct-pattern count.
  • Training on the larger ANERcorp + ACE 2005 mix increased recall at a small precision cost, giving a better overall F-measure.
  • For Experiment 2, ANERcorp person names in the test set were re-annotated as POL, ART, SPORT or Other.
    • Only new trigger lists were needed to handle these subtypes, and the results were comparable to the standard types.
    • Sportspersons had lower recall.

Interesting ideas

  • I1. Selective lexicalisation: a pattern keeps literal words only where they carry type signal (trigger nouns, verbs, prepositions) and reduces everything else to coarse POS, balancing generality and precision.
  • I2. Each full pattern is also split into a left-anchored and a right-anchored half, so a name can be caught even when only one side of its context matches.
  • I3. Candidates are ranked by the number of distinct patterns that found them, which is a diversity-of-evidence score, and this beat PMI.
  • I4. New fine-grained categories come almost for free by swapping in new trigger-word lists, while the corpus and algorithm stay unchanged.
  • I5. Patterns are matched against the POS-tagged corpus as-is, tolerating tagger errors where multiword names get NN/JJ tags instead of proper-noun tags.

For Mova

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

  • I1 (world): entity-type cues for the knowledge layer could be mined as POS skeletons anchored by a few indicator lemmas, giving readable rules instead of opaque features.
  • I2 (learn): self-induced repair rules could be stored in both full and one-sided variants, with a check of whether the shorter variant keeps precision on negative controls.
  • I3 (learn): silver annotations could be weighted by the count of independent witnesses (different parsers or rules) that agree, rather than by a single association score.
  • I4 (frames): verb-trigger lists like "elect" or "vote" could bias role or agent-type guesses (politician, athlete) in frames without retraining the parser.
  • I5 (tokens): when tagging proper-name spans, Mova could accept multi-token NN/JJ runs in name-like contexts as possible names, instead of trusting NNP tags alone.

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 EssexColchester · United Kingdom34 works · 3,537 citations

 Fields, topics and concepts

📚 Shelves 4The 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 works

🧭 Topics 2Research subjects and languages the work is about.Semitic languages · 2,908 worksNamed entity recognition · 200 works

Cite

Cite as Maha Althobaiti, Udo Kruschwitz and Massimo Poesio. 2013. A Semi-supervised Learning Approach to Arabic Named Entity Recognition. Proceedings of the International Conference Recent Advances in Natural Language Processing RANLP 2013.

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

@inproceedings{althobaiti2013semi,
  title = {{A Semi-supervised Learning Approach to Arabic Named Entity Recognition}},
  author = {Maha Althobaiti, Udo Kruschwitz and Massimo Poesio},
  booktitle = {Proceedings of the International Conference Recent Advances in Natural Language Processing RANLP 2013},
  year = {2013},
  url = {https://aclanthology.org/R13-1005/}
}

References

25 references, 7 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Sherief Abdallah, Khaled Shaalan, and Muhammad Shoaib. 2012. Integrating rule-based system with classification for arabic named entity recognition. In Computational Linguistics and Intelligent Text Processing, pages 311–322. Springer. find
  2. Samir AbdelRahman, Mohamed Elarnaoty, Marwa Magdy, and Aly Fahmy. 2010. Integrated machine learning techniques for arabic named entity recognition. IJCSI, 7:27–36. find
  3. Ahmed Abdul-Hamid and Kareem Darwish. 2010. Simplified feature set for arabic named entity recognition. In Proceedings of the 2010 Named Entities Workshop, pages 110–115. Association for Computational Linguistics.
  4. Maha Althobaiti, Udo Kruschwitz, and Massimo Poesio. 2012. Identifying named entities on a university intranet. In Computer Science and Electronic Engineering Conference (CEEC), 2012 4th, pages 94–99. IEEE. find
  5. Marco Baroni, Brian Murphy, Eduard Barbu, and Massimo Poesio. 2010. Strudel: A corpus-based semantic model based on properties and types. Cognitive Science, 34(2):222–254. ↗
  6. Yassine Benajiba and Paolo Rosso. 2007. Anersys 2.0: Conquering the ner task for the arabic language by combining the maximum entropy with pos-tag information. In IICAI, pages 1814–1823. find
  7. Yassine Benajiba and Paolo Rosso. 2008. Arabic named entity recognition using conditional random fields. In Proc. of Workshop on HLT & NLP within the Arabic World, LREC, volume 8, pages 143–153. find
  8. Yassine Benajiba, Paolo Rosso, and José Miguel Benedı́ruiz. 2007. Anersys: An arabic named entity recognition system based on maximum entropy. In Computational Linguistics and Intelligent Text Processing, pages 143–153. Springer. find
  9. Yassine Benajiba, Mona Diab, Paolo Rosso, et al. 2008. Arabic named entity recognition: An svm-based approach. In Proceedings of 2008 Arab International Conference on Information Technology (ACIT), pages 16–18. ↗
  10. Mona Diab. 2009. Second generation amira tools for arabic processing: Fast and robust tokenization, pos tagging, and base phrase chunking. In 2nd International Conference on Arabic Language Resources and Tools. find
  11. 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. Artificial Intelligence, 165(1):91–134. ↗
  12. Ralph Grishman and Beth Sundheim. 1996. Message understanding conference-6: A brief history. In COLING, volume 96, pages 466–471.
  13. Nizar Habash and Fatiha Sadat. 2006. Arabic preprocessing schemes for statistical machine translation. In NAACL. ACL.
  14. Shereen Khoja and Roger Garside. 1999. Stemming arabic text. Lancaster, UK, Computing Department, Lancaster University. find
  15. Wenhui Liao and Sriharsha Veeramachaneni. 2009. A simple semi-supervised algorithm for named entity recognition. In Proceedings of the NAACL HLT 2009 Workshop on Semi-Supervised Learning for Natural Language Processing, pages 58–65. Association for Computational Linguistics. find
  16. Xiaohua Liu, Shaodian Zhang, Furu Wei, and Ming Zhou. 2011. Recognizing named entities in tweets. In ACL, pages 359–367.
  17. Slim Mesfar. 2007. Named entity recognition for arabic using syntactic grammars. In Natural Language Processing and Information Systems, pages 305–316. Springer. Behrang Mohit, Nathan Schneider, Rishav Bhowmick, Kemal Oflazer, and Noah A Smith. Recall-oriented learning of named entities in arabic wikipedia. In Proceedings of the 13th Conference of the European Chapter of the Association for Computational Linguistics, pages 162–173. Association for Computational Linguistics. find
  18. David Nadeau and Satoshi Sekine. 2007. A survey of named entity recognition and classification. Lingvisticae Investigationes, 30(1):3–26. ↗
  19. David Nadeau, Peter Turney, and Stan Matwin. 2006. Unsupervised named-entity recognition: Generating gazetteers and resolving ambiguity. Advances in Artificial Intelligence. find
  20. Mai Oudah and Khaled F Shaalan. 2012. A pipeline arabic named entity recognition using a hybrid approach. In COLING, pages 2159–2176.
  21. Marius Pasca, Dekang Lin, Jeffrey Bigham, Andrei Lifchits, and Alpa Jain. 2006. Organizing and searching the world wide web of facts-step one: the one-million fact extraction challenge. In AAAI, volume 6, pages 1400–1405. find
  22. Ellen Riloff and Rosie Jones. 1999. Learning dictionaries for information extraction by multi-level bootstrapping. In AAAI, pages 474–479. find
  23. Khaled Shaalan and Hafsa Raza. 2007. Person name entity recognition for arabic. In Proceedings of the 2007 Workshop on Computational Approaches to Semitic Languages: Common Issues and Resources, pages 17–24. Association for Computational Linguistics.
  24. Khaled Shaalan and Hafsa Raza. 2009. Nera: Named 2012. entity recognition for arabic. Journal of the American Society for Information Science and Technology, 60(8):1652–1663. find
  25. Peter D Turney, Patrick Pantel, et al. 2010. From frequency to meaning: Vector space models of semantics. Journal of artificial intelligence research, 37(1):141–188. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/althobaiti-2013-semi-supervised-learning-approach-arabic/