Fixing the Infix: Unsupervised Discovery of Root-and-Pattern Morphology

What it is about

The authors learn root-and-pattern (templatic) morphology from raw text, for Arabic, Hebrew and Amharic, without labelled data. The method has two parts. An orthographic model proposes candidate rules: concatenative rules as sets of affixes added or deleted at one edge, and root-and-pattern rules as templates in which the three root consonants keep their order (for example C1AC2iC3 for an agent noun). A semantic model then keeps only the rules whose word pairs have the same difference vector in 64-dimensional Polyglot embeddings (for example kataba to maktab shares the vector difference that la'iba to mal'ab does). The discovered rules are used in JZR, an unsupervised root extractor. On 1,200 Arabic words JZR scores 51.88% against 61.63% for the engineered ISRI extractor, and 45.63% when limited to concatenative rules. A native speaker judged the top 30 rules in each language correct.

Ideas

• I1. A template list for Hebrew verbs and nouns in The Noun, the Adjective and the Numeral and The verb, checked by the rule of the paper: a candidate template is kept only when its pairs agree in a second, independent measure (for Mova, the lemma pairs of macula dict, not embeddings).
• I2. A root extractor for tools/hbo: they report that dropping the non-concatenative rules costs 11.5% relative, so the test is "with and without templates" on the same Hebrew words, reported as two numbers.
• I3. The 1,200-word sample with an engineered baseline (ISRI) is a model for a Hebrew root test set: take a sample from treebanks/hbo/ptnk, compare the root against a hand-checked root list (MACULA gives lemmas, not roots), and report the baseline beside the new method.

For Mova

I1 goes into The verb (templates) and The noun in Biblical Aramaic (Aramaic nouns, if the same rule holds). I2 is a measure for the root step of tools/hbo. I3 is a test set. The embedding-based validation is not a Mova method (Mova is deterministic and does not use vectors for rules); the transferable part is the rule "a template is valid only if its pairs agree", checked with a table.

Cite

The paper itself, not this card. BibTeX from the library's record of the work. arXiv 1702.02211

@misc{sakakini2017fixing,
  title = {{Fixing the Infix: Unsupervised Discovery of Root-and-Pattern Morphology}},
  author = {Sakakini, Tarek and Bhat, Suma and Viswanath, Pramod},
  year = {2017},
  eprint = {1702.02211},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1702.02211}
}

References

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

1. Mohamad Ababneh, Riyad Al-Shalabi, Ghassan Kanaan, and Alaa Al-Nobani. 2012. Building an effective rule-based light stemmer for arabic language to improve search effectiveness. International Arab Journal of Information Technology (IAJIT) 9(4). find
2. Mikko Kurimo, Sami Virpioja, Ville Al-Serhan and A. Ayesh. 2006. find
3. Turunen, and Krista Lagus. 2010. A triliteral word roots extraction using neural network for arabic. Morpho challenge competition 2005–2010: Evaluations and results. In 2006 International Conference on Com-In Proceedings of the 11th Meeting of the puter Engineering and Systems. pages 436–440. ACL Special Interest Group on Computa-https://doi.org/10.1109/ICCES.2006.320487. tional Morphology and Phonology. Association for Computational Linguistics, Stroudsburg, Riyad Al-Shalabi and Martha Evens. 1998. PA, USA, SIGMORPHON ’10, pages 87–95. A computational morphology system for arabic. http://dl.acm.org/citation.cfm?id=1870478.1870489. In Proceedings of the Workshop on Computational Approaches to Semitic Languages. ↗
4. Klaus Lagally. 1992. Arabtextypesetting arabic with Association for Computational Linguistics, vowels and ligatures. EuroTEX 92:153–172. Stroudsburg, PA, USA, Semitic ’98, pages 66–72. Leah S Larkey, Lisa Ballesteros, and Margaret E Con-http://dl.acm.org/citation.cfm?id=1621753.1621765. nell. 2002. Improving stemming for arabic information retrieval: light stemming and co-occurrence Yasir Alhanini, Mohd Juzaiddin Ab Aziz, et al. analysis. In Proceedings of the 25th annual inter-2011. The enhancement of arabic stemming by us-national ACM SIGIR conference on Research and ing light stemming and dictionary-based stemming. development in information retrieval. ACM, pages Journal of Software Engineering and Applications 275–282. 4(09):522. ↗
5. H. Yoong Keok Lee, Aria Haghighi, and Regina Barzi-Mohammed Aljlayl and Ophir Frieder. 2002. 2011. Modeling syntactic context improves On arabic search: Improving the retrieval effectiveness via lay. a light stemming approach. morphological segmentation. In Proceedings of In Proceedings of the Eleventh International Confer-the Fifteenth Conference on Computational Natural ence on Information and Knowledge Management. Language Learning. Association for Computational ACM, New York, NY, USA, CIKM ’02, pages Linguistics, pages 1–9. 340–347. https://doi.org/10.1145/584792.584848. ↗
6. Mohammed Attia, Ayah Zirikly, and Mona Diab. 2016. The power of language music: Arabic lemmatization through patterns. COLING 2016 page 40. find
7. Tim Buckwalter. 2004. Buckwalter arabic morphological analyzer version 2.0. linguistic data consortium, university of pennsylvania, 2002. ldc cat alog no.: Ldc2004l02. Technical report, ISBN 1-58563-324-0. find
8. Mathias Creutz and Krista Lagus. 2005. Unsupervised morpheme segmentation and morphology induction from text corpora using Morfessor 1.0. find
9. Samhaa R El-Beltagy and Ahmed Rafea. 2011. An accuracy-enhanced light stemmer for arabic text. ACM Transactions on Speech and Language Processing (TSLP) 7(2):2. find
10. John Goldsmith. 2000. Linguistica: An automatic morphological analyzer. In Proceedings of 36th meeting of the Chicago Linguistic Society. find
11. Joseph H Greenberg. 1950. The patterning of root morphemes in semitic. Word 6(2):162–181. find
12. Shereen Khoja and Roger Garside. 1999. Stemming arabic text. Lancaster, UK, Computing Department, Lancaster University . find
13. Tuomo Korenius, Jorma Laurikkala, Kalervo Järvelin, and Martti Juhola. 2004. Stemming and lemmatization in the clustering of finnish text documents. In Proceedings of the thirteenth ACM international conference on Information and knowledge management. ACM, pages 625–633. find
14. Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013a. Efficient estimation of word representations in vector space. arXiv preprint arXiv:1301.3781 . ↗
15. Tomas Mikolov, Wen-tau Yih, and Geoffrey Zweig. 2013b. Linguistic regularities in continuous space word representations. In Hlt-naacl. volume 13, pages 746–751. find
16. Arfath Pasha, Mohamed Al-Badrashiny, Mona T Diab, Ahmed El Kholy, Ramy Eskander, Nizar Habash, Manoj Pooleery, Owen Rambow, and Ryan Roth. 2014. Madamira: A fast, comprehensive tool for morphological analysis and disambiguation of arabic. In LREC. volume 14, pages 1094–1101. find our card
17. Jeffrey Pennington, Richard Socher, and Christopher D Manning. 2014. Glove: Global vectors for word representation. In EMNLP. volume 14, pages 1532– 43. find
18. Djamé Seddah, Grzegorz Chrupała, Özlem Çetinoğlu, Josef Van Genabith, and Marie Candito. 2010. Lemmatization and lexicalized statistical parsing of morphologically rich languages: the case of french. In Proceedings of the NAACL HLT 2010 First Workshop on Statistical Parsing of Morphologically-Rich Languages. Association for Computational Linguistics, pages 85–93. find
19. Nasredine Semmar, Meriama Laib, and Christian Fluhr. 2006. Using stemming in morphological analysis to improve arabic information retrieval. verbum ex machina page 317. find
20. Dorothy Carla Siegel. 1974. Topics in English morphology.. Ph.D. thesis, Massachusetts Institute of Technology. find
21. Lucie Skorkovská. 2012. Application of lemmatization and summarization methods in topic identification module for large scale language modeling data filtering. In International Conference on Text, Speech and Dialogue. Springer, pages 191–198. find
22. Radu Soricut and Franz Och. 2015. Unsupervised morphology induction using word embeddings. In Proc. NAACL. find
23. Kazem Taghva, Rania Elkhoury, and Jeffrey S Coombs. 2005. Arabic stemming without a root dictionary. In ITCC (1). pages 152–157. find

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/langstream-2017-sakakini-fixing-infix-unsupervised-discovery-root/