Elamur Research⌕
☰
home / papers / 2014 / Linking Databases using Matched Arabic Names

Linking Databases using Matched Arabic Names

Paper · 2014 · Tarek El-Shishtawy

Summary

The paper addresses matching personal names written in Arabic so that records about the same person can be joined across heterogeneous databases. Arabic names are hard for standard matchers for several reasons. Their tokens look very similar typographically. Their component order is fixed. People often drop middle tokens or common, low-information tokens. Purely character-based metrics miss token-level changes, and bag-of-words methods such as Soft TFIDF ignore order. The author therefore proposes a hybrid sequential algorithm that runs Levenshtein-style edit distance over name tokens rather than characters. Each edit operation is weighted by the token's position in the name, its frequency in a name frequency table, and its character-level similarity, subject to a threshold θ. The method was tested on a large Arabic dataset under several synthetic error scenarios and compared with classical metrics. It raised the worst-case success rate from about 69% (Soft TFIDF) to about 80%, with a best case of 99.67%. In a real university deployment it linked 98.5% of records automatically, which makes it practical for data integration.

Key points

  • The core method extends Levenshtein distance to the token level, using a position weight β and a character similarity threshold θ. Token frequency also adjusts the cost of each edit.
  • Results are reported as lower and upper success limits across error types:
    • The proposed method spans about 80% to 99.67%.
    • Soft TFIDF has a minimum of 69%.
    • Basic Levenshtein ranges from 16% to 100%, reaching its maximum on one- or two-character omissions.
  • The minimum success level was measured under omission of the second and third tokens. "Success" means the correct name is the top-scoring candidate.
  • In a real deployment for the Egyptian Universities Portals Project (EUP, funded by MOHE), the method linked a Benha University staff table (~3500 people) to a training database (2200 records).
    • Exact matching found only 14% of the training records.
    • The algorithm linked 98.5% with no manual work.
  • The approach can be ported to other languages with similar naming conventions by replacing the Arabic name frequency table.

Interesting ideas

  • I1. Edit distance can be computed over whole tokens instead of characters, with each operation's cost depending on where the token sits and how common it is. This keeps word order meaningful while tolerating dropped or inserted parts.
  • I2. A single threshold θ moves the matcher between behaviour like a pure token matcher (θ = 1) and behaviour like a character-level matcher (θ = 0). This makes the granularity an explicit, tunable dial.
  • I3. Instead of one average score, the paper reports lower and upper success limits across deliberately distorted inputs, one distortion type at a time. A method with a narrow, high range is preferred because it is more robust.
  • I4. Success is defined as the correct answer ranking first, because the goal is a fully automatic pipeline rather than a shortlist for humans.
  • I5. Language-specific knowledge is isolated in a swappable frequency table, so the algorithm itself stays general.

For Mova

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

  • I1 (world): Matching entity mentions such as "Sir John Smith" against "John Smith" in the world of a text could use position- and frequency-weighted token edits to merge entities without ignoring order.
  • I2 (frames): Coreference of named mentions could expose one such dial, adjusted per genre (more typos versus more omitted name parts), and log the chosen value for explainability.
  • I3 (learn): The evaluation could apply controlled corruption types (dropped tokens, misspellings) to inputs and report worst-case and best-case accuracy, exposing brittle components that averages hide.
  • I4 (answer): Readers could be scored on top-1 correctness when answers feed later steps automatically, which keeps the metric aligned with the actual use.
  • I5 (world): Name and token frequency statistics could live as a separate compiled knowledge resource, so matching logic can be reused for other name systems or domains.

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 6The 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 worksDecision Sciences · 249 worksManagement Science and Operations Research · 134 works

🧭 Topics 1Research subjects and languages the work is about.Semitic languages · 3,066 works

Cite

Cite as Tarek El-Shishtawy. 2014. Linking Databases using Matched Arabic Names. International Journal of Computational Linguistics & Chinese Language Processing, Volume 19, Number 1, March 2014.

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

@article{elshishtawy2014linking,
  title = {{Linking Databases using Matched Arabic Names}},
  author = {Tarek El-Shishtawy},
  journal = {International Journal of Computational Linguistics \& Chinese Language Processing, Volume 19, Number 1, March 2014},
  year = {2014},
  url = {https://aclanthology.org/O14-2003/}
}

References

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

  1. Bilenko, M., Mooney, R., Cohen, W., Ravikumar, P., & Fienberg, S. (2003). Adaptive name matching in information integration. Intelligent Systems, IEEE , 18 (5), 16-23. ↗
  2. Bilenko, M., & Mooney, R. (2003). Adaptive duplicate detection using learnable string similarity measures. In Proceedings of the ninth ACM SIGKDD international conference on Knowledge discovery and data mining (2003) , 39-48. find
  3. Borgman, C. L., & Siegfried, S. L. (1999). Getty's Synoname™ and its cousins: A survey of applications of personal name matching algorithms. Journal of the American Society for Information Science , 43 (7), 459-476 find
  4. Camacho, D., Huerta, R., & Elkan, C. (2008). An Evolutionary Hybrid Distance for Duplicate String Matching. http://arantxa.ii.uam.es/~dcamacho/StringDistance/hybrid-distance.pdf . ↗
  5. Cohen, W., Ravikumar, P., & Fienberg, S. (2003). A comparison of string distance metrics for name-matching tasks. In Proceedings of the IJCAI-2003 Workshop on Information Integration on the Web (IIWeb-03) , 73-78. find
  6. Elmagarmid, A., Ipeirotis, P., & Verykios, V. (2007). Duplicate record detection: A survey. IEEE Transactions on Knowledge and Data Engineering , 19 (1), 1-16. ↗
  7. Gadd, T. N. (1990). PHONIX: The algorithm. Program: electronic library and information systems , 24 (4), 363-366. ↗
  8. Hall, P., & Dowling, G. (1980). Approximate string matching. ACM Computing Surveys (CSUR) , 12 (4), 381-402. ↗
  9. Holmes, D., & McCabe, M. (2002). Improving precision and recall for soundex retrieval. In Information Technology: Proceedings. International Conference on Coding and Computing 2002 , 22-26. find
  10. Jaro, M. (1989). Advances in record-linkage methodology as applied to matching the 1985 census of Tampa, Florida. Journal of the American Statistical Association , 84 (406), 414-420. ↗
  11. Jurafsky, D., Martin, J., & Kehler, A. (2002). Speech and language processing: an introduction to natural language processing, computational linguistics and speech recognition (Vol. 2). Prentice Hall. find
  12. Keskustalo, H., Pirkola, A., Visala, K., Leppänen, E., & Järvelin, K. (2003). Non-adjacent digrams improve matching of cross-lingual spelling variants. In String Processing and Information Retrieval, Lecture Notes in Computer Science Volume 2857, 252-265. Springer Berlin/Heidelberg. find
  13. Monge, A., & Elkan, C. (1996). The field matching problem: Algorithms and applications. In Proceedings of the second international Conference on Knowledge Discovery and Data Mining , 267-270. find
  14. Moreau, E., Yvon, F., & Cappé, O. (2008). Robust similarity measures for named entities matching. In Proceedings of the 22nd International Conference on Computational Linguistics , 1 , 593-600. find
  15. Patman, F., & Thompson, P. (2003). Names: A new frontier in text mining. Intelligence and Security Informatics, Lecture Notes in Computer Science Volume 2665, 960-960. find
  16. Smith, T., & Waterman, M. (1981). ªIdentification of Common Molecular Subsequences. º J. Molecular Biology , 147 , 195-197. find
  17. Snae, C. (2007). A comparison and analysis of name matching algorithms. International Journal of Applied Science, Engineering and Technology , 4 (1), 252-257. ↗
  18. Waterman, M., Smith, T., & Beyer, W. (1976). Some biological sequence metrics. Advances in Mathematics , 20 (3), 367-387. ↗
  19. Winkler, W. (2002). Methods for record linkage and Bayesian networks. Technical report, Statistical Research Division , (2002). US Census Bureau, Washington, DC. find
  20. Winkler, W. (2006). Overview of record linkage and current research directions. In Bureau of the Census. (2006). find
  21. Xiao, C., Wang, W., Lin, X., Yu, J., & Wang, G. (2011). Efficient similarity joins for near-duplicate detection. ACM Transactions on Database Systems (TODS) , 36 (3), 15. ↗

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/shishtawy-2014-linking-databases-matched-arabic-names/