Elamur Research⌕
☰
home / papers / 2022 / Current Trends and Approaches in Synonyms Extraction: Potential Adaptation to Arabic

Current Trends and Approaches in Synonyms Extraction: Potential Adaptation to Arabic

Summary

This survey reviews ways to obtain synonyms automatically rather than by hand. Hand-built resources such as Princeton WordNet are slow and costly to produce, which is why most WordNets for other languages, Arabic included, remain small. The authors sort prior work into four families. The first infers synonyms from translation graphs built out of bilingual dictionaries. The second derives new translation pairs by pivoting through a third language. The third assembles WordNet synsets by exploring synonymy and translation graphs. The fourth mines similar words from corpora using word embeddings and, more recently, BERT-based lexical substitution. For each family the paper describes the method, the data and the evaluation used, and then compares the families. A recurring finding is that graph-based methods can be precise, but their output is only as good as the input dictionaries. Polysemy produces spurious links, and no standard benchmark exists, so most work is evaluated against existing WordNets or by manual checking. The survey closes by pointing to these techniques as ways to grow Arabic synonym resources.

Key points

  • Arabic translation-graph method [6]. Synonym sets are found from cycles in an undirected Arabic–English translation graph, then merged by taking the union of sets that share translations. To test it, the authors flattened the Arabic WordNet (AWN) into a bilingual dictionary and rebuilt it. Agreement with the original, measured by cosine similarity, was 82%.
  • English–Italian thesaurus [8]. Missing synonyms were found using Cyclic and Quasi-Cyclic (CQC) paths, explored with depth-first search and scored so that shorter paths count more. It reached about 80% accuracy at the sixth level and was evaluated on the TOEFL dataset.
  • Translation-pair discovery over Apertium dictionaries [9]. Cycle-based, path-based and multi-way NMT methods were combined. Recall was very low, with precision between 25% and 75%.
  • Apertium RDF graph with 22 dictionaries [10]. Translations were scored by dense cycles, with exploration capped at level 7. After the English–Spanish link was deleted, the method recovered 72% of it. It also matched 73% of a converted Wiktionary English–French set.
  • Synset construction from multiple Wiktionaries [11]. Clustering, greedy expansion and supervised classification were applied to build Slovenian, Persian, German and Russian WordNets. Accuracy ranged from 20% to 88%, and exceeded 90% when strict thresholds were used.

Interesting ideas

  • I1. Testing an induction method by flattening a trusted resource (here AWN) and checking how much the algorithm rebuilds gives a gold standard at no annotation cost. — For Mova (learn): we could remove part of our compiled verb classes or ontology links and check whether our induction rules regenerate them, which acts as a built-in positive control.
  • I2. A translation round trip that returns to the starting word is treated as evidence that every word on the loop shares a meaning. Overlapping candidate sets are then merged by union instead of being ranked. — For Mova (world): within the WordNet/Wikidata knowledge graph, closed equivalence loops could license merging entries into one concept, and each merge would stay explainable as the loop that justified it.
  • I3. Path scoring that prefers short paths, combined with a hard depth limit, keeps graph walks cheap and limits drift caused by polysemous hops. — For Mova (world): when inferring relations through the knowledge graph, we can cap hop count and down-weight long chains, and show the winning path as the explanation.
  • I4. The threshold trade-off is made explicit: a strict setting yields small, accurate synsets, while a loose setting yields larger ones that need a follow-up corpus filter. The threshold must be tuned on validation data kept apart from the test set. — For Mova (learn): silver-data agreement thresholds could be tuned the same way on a held-out split, with a stricter tier for training and a looser tier sent to filtering.
  • I5. Errors caused by polysemy can be reduced by requiring agreement through several pivot languages rather than a single one. — For Mova (back): when translating through the meta-world, a lexical choice should be accepted only if more than one independent route supports it, so that one ambiguous bridge word cannot decide it.

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: Linguistics & languages

Topics: Word sense disambiguation

Cited by

5 papers of our library cite it (of the 140884 whose bibliographies we read); 1 of them have a page here, 4 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.1109/icit52682.2021.9491713 · arXiv 2205.10412

@misc{naserkarajah2022current,
  title = {{Current Trends and Approaches in Synonyms Extraction: Potential Adaptation to Arabic}},
  author = {Naser-Karajah, Eman and Arman, Nabil and Jarrar, Mustafa},
  year = {2022},
  doi = {10.1109/icit52682.2021.9491713},
  eprint = {2205.10412},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.1109/icit52682.2021.9491713}
}

References

38 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. S.J Tshephisho, and T. B. Mokgonyane. (2021) "Practical approach on implementation of WordNets for South African languages.". find
  2. G. A. Miller, R. Beckwith, C. Fellbaum, D.Gross, Katherine and J. Miller, (1990) Introduction to WordNet: An On-line Lexical Database, International Journal of Lexicography, Volume 3, Issue 4, Winter 1990, Pages 235–244. find
  3. S. Neale. (2018).A survey on automatically-constructed WordNets and their evaluation: Lexical and word embedding-based approaches. In Proceedings of the Eleventh International Conference on Language Resources and Evaluation. find
  4. N. Boukhatem, (2014) The Arabic Natural Language Processing: Introduction and Challenges. International Journal of English Language & Translation Studies, 2(3): p. 106-112. find
  5. A.Farghaly, and K. Shaalan (2009). Arabic natural language processing: Challenges and solutions. ACM Transactions on Asian Language Information Processing (TALIP) find
  6. M. Jarrar, E. Karajah, M. Khalifa and K.Shaalan. (2020) Extracting Synonyms from Bilingual Dictionaries. Proceedings of the 11th Global WordNet Conference. find
  7. N. Mohammed, (2020). Extracting word synonyms from text using neural approaches. Int. Arab J. Inf. Technol., 17(1), 45-51. find
  8. T. Flati, R. Navigli (2012). The CQC algorithm: Cycling in graphs to semantically enrich and enhance a bilingual dictionary. Journal of Artificial Intelligence Research, 43, 135-171 find
  9. D. Torregrosa, M. Arcan, S. Ahmadi and J. P. McCrae (2019). Tiad 2019 shared task: Leveraging knowledge graphs with neural machine translation for automatic multilingual dictionary generation. Translation Inference Across Dictionaries. find
  10. M. Villegas, M. Melero, J.Gracia and N. Bel. (2016, May). Leveraging RDF graphs for crossing multiple bilingual dictionaries. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC'16) (pp. 868-876). find
  11. G.Ercan and F.Haziyevb (2019). Synset expansion on translation graph for automatic WordNet construction. Information Processing & Management, 56(1), 130-150. find
  12. M. Jarrar (2019). The Arabic ontology–an Arabic WordNet with ontologically clean content. Applied Ontology, (Preprint), 1-26. find our card
  13. H.WU and M. ZHOU (2003, July). Optimizing synonym extraction using monolingual and bilingual resources. In Proceedings of the second international workshop on Paraphrasing (pp. 72-79). find
  14. K. N.Lam, F.Al Tarouti and J. Kalita (2014, June). Automatically constructing WordNet synsets. In Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers) (pp. 106-111). find
  15. F. Al Tarouti and J. Kalita (2016, June). Enhancing automatic WordNet construction using word embeddings. In Proceedings of the Workshop on Multilingual and Cross-lingual Methods in NLP (pp. 30-34). find
  16. M. Khodak, A. Risteski, C. Fellbaum and S. Arora (2017, April). Automated WordNet construction using word embeddings. In Proceedings of the 1st Workshop on Sense, Concept and Entity Representations and their Applications (pp. 12-23). find
  17. S. Arora, Y Li, Y. Liang, T. Ma and A. Risteski (2016). A latent variable model approach to pmi-based word embeddings. Transactions of the Association for Computational Linguistics, 4, 385-399.(2016) find our card
  18. M. Jarrar and H. Amayreh (2019, June). An Arabic-multilingual database with a lexicographic search engine. In International Conference on Applications of Natural Language to Information Systems (pp. 234-246). Springer, Cham. find
  19. M. Jarrar, F. Zaraket, R. Asia and H. Amayreh (2018). Diacritic-based matching of Arabic words. ACM Transactions on Asian and Low-Resource Language Information Processing (TALLIP), 18(2), 1-21. find
  20. R.N. Al-Matham and H. S. Al-Khalifa.(2021) .SynoExtractor: A Novel Pipeline for Arabic Synonym Extraction Using Word2Vec Word Embeddings.Complexity. find
  21. M.AlMaayah, M.Sawalha, and M. A. Abushariah. (2016). Towards an automatic extraction of synonyms for Quranic Arabic WordNet. International Journal of Speech Technology 19, no. 2 (2016): 177-189. find
  22. W.Zhou, T.Ge, K.Xu, F.Wei, and M. Zhou (2019). "BERT-based lexical substitution." In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pp. 3368-3373. find
  23. D.McCarthy, Diana, and R. Navigli (2009). "The English lexical substitution task." Language resources and evaluation 43, no. 2 (2009): 139-159 . find
  24. J.Devlin, M.Chang, K.Lee, and K.Toutanova (2018). Bert: Pre-training of deep bidirectional transformers for language understanding. arXiv preprint arXiv:1810.04805. ↗
  25. K. Tanaka and K. Umemura (1994). construction of a bilingual dictionary intermediated by a third language. In COLING, pages 297–303. find
  26. L. T. Lim, B. Ranaivo-Malançon, and E. K. Tang (2011). Low cost construction of a multilingual lexicon from bilingual lists. Polibits, 43:45–51,(2011). find
  27. S.Soderland, O. Etzioni, D. S. Weld, M. Skinner, and J. Bilmes. Compiling a massive, multilingual dictionary via probabilistic inference." In Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International Joint Conference on Natural Language Processing of the AFNLP, pp. 262-270. 2009 . find
  28. D.Ustalov, A.Panchenko, and Chris Biemann (2017). Watset: Automatic induction of synsets from a graph of synonyms." arXiv preprint arXiv:1704.07157. ↗
  29. D.Ustalov, M. Chernoskutov, C.Biemann, and A.Panchenko (2017). Fighting with the sparsity of synonymy dictionaries for automatic synset induction. In International Conference on Analysis of Images, Social Networks and Texts,( pp. 94-105). Springer, Cham. find
  30. R.Navigli, and S.P.Ponzetto (2012). BabelNet: The automatic construction, evaluation and application of a wide-coverage multilingual semantic network.Artificial intelligence 193: 217-250. find
  31. I.Gurevych, J.Eckle-Kohler, S.Hartmann, M.Matuschek, C.M. Meyer, and C. Wirth (2012). UBY-a large-scale unified lexical-semantic resource based on LMF. In Proceedings of the 13th Conference of the European Chapter of the Association for Computational Linguistics, pp. 580-590. find
  32. M.Helou, Abu, M. Palmonari, and M. Jarrar (2016). Effectiveness of automatic translations for cross-lingual ontology mapping." Journal of Artificial Intelligence Research 55: 165-208. find
  33. Al-Hajj, M., Jarrar, M., (2021). LU-BZU at SemEval-2021 Task 2: Word2Vec and Lemma2Vec performance in Arabic Word-in-Context disambiguation. In Proceedings – the 11th Workshop on Semantic Evaluation (SemEval2021). find
  34. Darwish, K., Habash, N., Abbas, M., Al-Khalifa, H., Al-Natsheh, H., Bouamor, H., Bouzoubaa, K., Cavalli-Sforza, V., El-Beltagy, S., El-Hajj, W., Jarrar, M., Mubarak, H. (2021). A Panoramic Survey of Natural Language Processing in the Arab World. Communications of the ACM, April 2021, Vol. 64 No. 4, Pages 72-81 find
  35. Jarrar, M. (2011): Building A Formal Arabic Ontology (Invited Paper). In Proceedings – Experts Meeting on Arabic Ontologies and Semantic Networks, Tunis. ALECSO, Arab League. find
  36. Jarrar, M., Amayreh, H., & McCrae, J. (2019): Representing Arabic Lexicons in Lemon – a Preliminary Study. In Proceedings – 2nd Conference on Language, Data and Knowledge, Leipzig, Germany. CEUR-WS (vol. 2402, pp. 29-33). find
  37. Jarrar, M. (2020). Digitization of Arabic Lexicons. Arabic Language Status Report. UAE Ministry of Culture and Youth. Pages 214-2017. find
  38. Alhafi, D., Deik, D., & Jarrar, M. (2019): Usability Evaluation of Lexicographic e-Services. In Proceedings – 2019 IEEE/ACS 16th International Conference on Computer Systems and Applications, Abu Dhabi (pp.1-7). IEEE. find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/naser-2022-karajah-current-trends-approaches-synonyms-extraction/