Picking Up Where the Linguist Left Off: Mapping Morphology to Phonology through Learning the Residuals

What it is about

The task is to produce the spoken form of Cairene Egyptian Arabic words from their underlying morphological representation, using the Egyptian Colloquial Arabic Lexicon as gold. A linguist supplies six ordered morphophonological rewrite rules, plus syllabification and stress assignment, implemented as regular expressions. Because the rules miss some cases and the morphological input has errors, a character-level transformer is trained to correct the rules' output (the residual) rather than to map morphology to speech directly. On a learning curve from 100 to about 12,000 training words, the rules alone win in the lowest range (under about 1,000 words), while the residual model beats both the rules and a neural model trained without rules from about 600 to 1,000 examples on. The authors report 78.6% for the residual model, 73.2% for the rules and 65.9% for the plain transformer at 900 examples.

Ideas

• I1. Keep the linguist's rules as the deterministic first stage and add a small learned correction stage that only sees the rules' output and the error cases they leave. In Mova this is a second, separate pass: the rules stay as they are, and the correction table is learned and inspected, never run as a black box.
• I2. Measure each rule's contribution by applying the rules incrementally in their order and recording accuracy after each one (the paper's Table 1 layout). The same incremental table, with a per-rule count of fixed and broken forms, is a cheap ablation of any ordered rule set.
• I3. Sort the residual errors into classes (added gemination, dropped consonants, wrong vowel operations, and so on). The paper's error taxonomy can be used as a checklist for a Hebrew or Aramaic rule set, which will show where a rule is missing rather than where it is wrong.

For Mova

I2 is a measurement for any ordered rule set, in the spirit of the value ablations of tools/value. I1 and I3 belong to the phonology chapters: 3. Phonology (a grapheme-to-phoneme rule set for a Modern Hebrew reader) and Phonology of Biblical Hebrew (syllable-sensitive rules). The rules themselves go in those chapters, not in the code.

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.18653/v1/2024.arabicnlp-1.22

@inproceedings{khalifa2024picking,
  title = {{Picking Up Where the Linguist Left Off: Mapping Morphology to Phonology through Learning the Residuals}},
  author = {Khalifa, Salam and Qaddoumi, Abdelrahim and Broselow, Ellen and Rambow, Owen},
  booktitle = {Proceedings of the Second Arabic Natural Language Processing Conference},
  year = {2024},
  doi = {10.18653/v1/2024.arabicnlp-1.22},
  url = {https://doi.org/10.18653/v1/2024.arabicnlp-1.22}
}

References

18 references, 1 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Ernest T. Abdel-Massih, Zaki N. Abdel-Malek, and El-Said M. Badawi. 1979. A Reference Grammar of Egyptian Arabic. Georgetown University Press. find
2. Daud Atiyeh Abdo. 1969. Stress and Arabic phonology. University of Illinois at Urbana-Champaign. find
3. Mahasen Hasan Abu-Mansour. 1987. A nonlinear analysis of Arabic syllabic phonology, with special reference to Makkan. University of Florida. find
4. Fadi Biadsy, Nizar Habash, and Julia Hirschberg. 2009a. Improving the Arabic Pronunciation Dictionary for Phone and Word Recognition with Linguistically-Based Pronunciation Rules. In Proceedings of the Conference of the North American Chapter of the Association for Computational Linguistics (NAACL), pages 397–405, Boulder, Colorado. find
5. Fadi Biadsy, Julia Hirschberg, and Nizar Habash. 2009b. Spoken Arabic Dialect Identification Using Phonotactic Modeling. In Proceedings of the Workshop on Computational Approaches to Semitic Languages (CASL), pages 53–61, Athens, Greece. find
6. Houda Bouamor, Nizar Habash, Mohammad Salameh, Wajdi Zaghouani, Owen Rambow, Dana Abdulrahim, Ossama Obeid, Salam Khalifa, Fadhl Eryani, Alexander Erdmann, and Kemal Oflazer. 2018. The MADAR Arabic dialect corpus and lexicon. In Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018), Miyazaki, Japan. European Language Resources Association (ELRA). find
7. Mohamed Boudchiche, Azzeddine Mazroui, Mohamed Ould Abdallahi Ould Bebah, Abdelhak Lakhouaja, and Abderrahim Boudlal. 2017. AlKhalil Morpho Sys 2: A robust Arabic morpho-syntactic analyzer. Journal of King Saud University - Computer and Information Sciences, 29(2):141–146. find
8. Ellen Broselow. 1976. The Phonology of Egyptian Arabic. Ph.D. thesis. Copyright - Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works; Last updated -2023-07-26. find
9. Ellen Broselow. 2017. Syllable Structure in the Dialects of Arabic. The Routledge handbook of Arabic linguistics, pages 32–47. find
10. Noam Chomsky and Morris Halle. 1968. The Sound Pattern of English. Harper & Row New York. find
11. Nizar Habash, Ramy Eskander, and Abdelati Hawwari. 2012. A Morphological Analyzer for Egyptian Arabic. In Proceedings of the Workshop of the Special Interest Group on Computational Morphology and Phonology (SIGMORPHON), pages 1–9, Montréal, Canada. Nizar Habash, Salam Khalifa, Fadhl Eryani, Owen Rambow, Dana Abdulrahim, Alexander Erdmann, Reem Faraj, Wajdi Zaghouani, Houda Bouamor, Nasser Zalmout, Sara Hassan, Faisal Al shargi, Sakhar Alkhereyf, Basma Abdulkareem, Ramy Eskander, Mohammad Salameh, and Hind Saddiki. 2018. Unified Guidelines and Resources for Arabic Dialect Orthography. In Proceedings of the International Conference on Language Resources and Evaluation (LREC 2018), Miyazaki, Japan. find our card
12. Ghassan F Haddad. 1984. Problems and issues in the phonology of Lebanese Arabic. University of Illinois at Urbana-Champaign. find
13. Abdel Halim Hamid. 1984. A Descriptive Analysis of Sudanese Colloquial Arabic Phonology. Ph.D. thesis. Copyright - Database copyright ProQuest LLC; ProQuest does not claim copyright in the individual underlying works; Last updated - 2023-02-19. find
14. Larry M. Hyman. 2018. Why underlying representations? Journal of Linguistics, 54(3):591–610. find
15. Christian Khairallah, Salam Khalifa, Reham Marzouk, Mayar Nassar, and Nizar Habash. 2024. A large-scale open-source morphological analyzer for modern standard arabic. In Proceedings of the joint LREC-COLING conference. find
16. Salam Khalifa, Sara Hassan, and Nizar Habash. 2017. A morphological analyzer for Gulf Arabic verbs. In Proceedings of the Workshop for Arabic Natural Language Processing (WANLP), Valencia, Spain. find
17. Salam Khalifa, Jordan Kodner, and Owen Rambow. 2022. Towards learning Arabic morphophonology. In Proceedings of the The Seventh Arabic Natural Language Processing Workshop (WANLP), Abu Dhabi, United Arab Emirates. Association for Computational Linguistics. find
18. Salam Khalifa, Sarah Payne, Jordan Kodner, Ellen Broselow, and Owen Rambow. 2023. A cautious generalization goes a long way: Learning morphophonological rules. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 1793–1805, Toronto, Canada. Association for Computational Linguistics. Hanaa Kilany, Hassan Gadalla, Howaida Arram, Ashraf Yacoub, Alaa El-Habashi, and Cynthia McLemore. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/aclbulk-2024-khalifa-picking-up-where-linguist-left/