Elamur Research⌕
☰
home / papers / 2025 / Universal Dependencies for Sindhi

Universal Dependencies for Sindhi

Summary

Sindhi is an Indo-Aryan language with split ergativity, spoken by about 40 million people in Pakistan and India. Despite this, it has few resources for NLP. The authors created a new UD treebank for Sindhi and examined how to describe the language's complex phenomena in UD. These include case markers that themselves inflect, pronominal suffixes, intensifiers, and demonstrative pronouns that can function as determiners. To speed up annotation, silver trees were produced first. For this, Stanza was trained on a mixture of UD data from Hindi, Urdu, Marathi and Tamil with the Muril transformer. Even a few hundred annotated Sindhi sentences yielded a high-quality initial annotation. The authors also compiled a Sindhi text corpus larger than previous ones and trained static embeddings on it. From these components they built the first end-to-end annotation pipeline for Sindhi in Stanza. The work is worth knowing as an example of how to quickly obtain a UD resource for a low-resource language by leveraging closely related languages.

Key points

  • About 6000 sentences have been annotated. The dataset was prepared for UD releases 2.16 and 2.17. Part of the test set is drawn from 509 re-annotated sentences of the older Mazhar Dootio corpus.
  • Inter-annotator agreement on 200 silver sentences is 96% for UPOS and 91.4% for dependencies.
  • The cross-lingual model (Hindi, Urdu, Marathi, Tamil together with 500 Sindhi trees) achieves LAS 82.68 on the test set. Tamil, although Dravidian, also slightly improves the result.
  • On the test set, the parser reaches LAS ≈91.8, and NER (trained on SiNER) reaches entity F1 ≈88.1. These values were recovered from a table damaged during extraction, so they should be checked against the original.
  • Custom Sindhi transformers, both further-pretrained and trained from scratch, underperformed the original Muril-Large. In contrast, embeddings trained specifically on Sindhi outperformed multilingual ones.

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: Universal Dependencies

Cite

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

@inproceedings{bauer2025universal,
  title = {{Universal Dependencies for Sindhi}},
  author = {Bauer, John and Shah, Sakiina and Shaheer, Muhammad and Talpur, Mir Afza Ahmed and Sanjrani, Zubair and Qureshi, Sarwat and Pirzada, Shafi and Manning, Christopher D. and Rahman, Mutee U},
  booktitle = {Proceedings of the Eighth Workshop on Universal Dependencies (UDW, SyntaxFest 2025)},
  year = {2025},
  url = {https://aclanthology.org/2025.udw-1.11/}
}

References

38 references, 10 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Julien Abadji, Pedro Ortiz Suarez, Laurent Romary, and Benoît Sagot. 2022. Towards a Cleaner Document-Oriented Multilingual Crawled Corpus. arXiv e-prints, arXiv:2201.06642. find our card
  2. Alan Akbik, Duncan Blythe, and Roland Vollgraf. 2018. Contextual string embeddings for sequence labeling. In Proceedings of the 27th International Conference on Computational Linguistics, pages 1638– 1649, Santa Fe, New Mexico, USA. Association for Computational Linguistics. in Universal Dependencie… find our card
  3. Meesum Alam, Francis Tyers, Emily Hanink, and Sandra Kübler. 2024. Universal Dependencies for Saraiki. In Proceedings of the Joint Workshop on Multiword Expressions and Universal Dependencies (MWE-UD) @ LREC-COLING 2024, pages 188– 197, Torino, Italia. ELRA and ICCL. find our card
  4. Wazir Ali, Jay Kumar, Junyu Lu, and Zenglin Xu. 2020a. Word embedding based new corpus for low-resourced language: Sindhi. Preprint, arXiv:1911.12579. — [arXiv:1911.12579](https://arxiv.org/abs/1911.12579) find
  5. Wazir Ali, Junyu Lu, and Zenglin Xu. 2020b. SiNER: A large dataset for Sindhi named entity recognition. In Proceedings of the Twelfth Language Resources and Evaluation Conference, pages 2953–2961, Marseille, France. European Language Resources Association. find
  6. Wazir Ali, Zenglin Xu, and Jay Kumar. 2021. SiPOS: A benchmark dataset for Sindhi part-of-speech tagging. In Proceedings of the Student Research Workshop Associated with RANLP 2021, pages 22–30, Online. IN-COMA Ltd. find our card
  7. Aryaman Arora. 2022. Universal Dependencies for Punjabi. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 5705– 5711, Marseille, France. European Language Resources Association. find our card
  8. John Bauer, Chloé Kiddon, Eric Yeh, Alex Shan, and Christopher D. Manning. 2023. Semgrex and ssurgeon, searching and manipulating dependency graphs. In Proceedings of the 21st International Workshop on Treebanks and Linguistic Theories (TLT, GURT/SyntaxFest 2023), pages 67–73, Washington, D.C. Assoc… find
  9. Yevgeni Berzak, Jessica Kenney, Carolyn Spadine, Jing Xian Wang, Lucia Lam, Keiko Sophie Mori, Sebastian Garza, and Boris Katz. 2016. Universal Dependencies for learner English. In Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages… find our card
  10. Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Edouard Grave, Myle Ott, Luke Zettle-Unsuper-moyer, and Veselin Stoyanov. 2020. vised cross-lingual representation learning at scale. Preprint, arXiv:1911.02116. find
  11. Marie-Catherine de Marneffe, Christopher D. Manning, Joakim Nivre, and Daniel Zeman. 2021. Universal Dependencies. Computational Linguistics, 47(2):255–308. find our card
  12. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of deep bidirectional transformers for language understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies,… find
  13. Mazhar Dootio and Asim Wagan. 2017. Automatic stemming and lemmatization process for Sindhi text. JSSIR, 6:19–28. find
  14. Mazhar Ali Dootio and Asim Imdad Wagan. 2019. Syntactic parsing and supervised analysis of Sindhi text. Journal of King Saud University – Computer and Information Sciences, 31(1):105–112. find
  15. Luke Gessler and Amir Zeldes. 2022. MicroBERT: Effective training of low-resource monolingual BERTs through parameter reduction and multitask learning. In Proceedings of the 2nd Workshop on Multi-lingual Representation Learning (MRL), pages 86–99, Abu Dhabi, United Arab Emirates (Hybrid). Associatio… find
  16. Fadli Aulawi Al Ghiffari, Ika Alfina, and Kurni-awati Azizah. 2024. Cross-lingual transfer learning for Javanese dependency parsing. Preprint, arXiv:2401.12072. find
  17. Akshar Bharati, Dipti Misra Sharma, Lakshmi Bai, and Rajeev Sangal. 2008. Anncorra : Annotating corpora guidelines for pos and chunk annotation for Indian languages. Technical report, Language Technologies Research Center, IIIT Hyderabad. find
  18. Riyaz Ahmad Bhat, Rajesh Bhatt, Annahita Farudi, Prescott Klassen, Bhuvana Narasimhan, Martha Palmer, Owen Rambow, Dipti Misra Sharma, Ashwini Vaidya, Sri Ramagurumurthy Vishnu, and 1 others. 2017. The hindi/urdu treebank project. In Handbook of Linguistic Annotation. Springer Press. find
  19. Flavio Massimiliano Cecchini. 2021. Formae reformandae: for a reorganisation of verb form annotation Edouard Grave, Piotr Bojanowski, Prakhar Gupta, Armand Joulin, and Tomas Mikolov. 2018. Learning word vectors for 157 languages. In Proceedings of the International Conference on Language Resources a… find
  20. Johannes Heinecke. 2019. ConlluEditor: a fully graphical editor for universal dependencies treebank files. In Proceedings of the Third Workshop on Universal Dependencies (UDW, SyntaxFest 2019), pages 87–93, Paris, France. Association for Computational Linguistics. find
  21. Matthew Honnibal, Ines Montani, Sofie Van Lan-deghem, and Adriane Boyd. 2020. spaCy: Industrial-strength Natural Language Processing in Python. find
  22. Mohammed Khan, Priyam Mehta, Ananth Sankar, Umashankar Kumaravelan, Sumanth Doddapaneni, Suriyaprasaad B, Varun G, Sparsh Jain, Anoop Kunchukuttan, Pratyush Kumar, Raj Dabre, and Mitesh Khapra. 2024. IndicLLMSuite: A blueprint for creating pre-training and fine-tuning datasets for Indian languages.… find
  23. Loganathan Ramasamy and Zdeněk Žabokrtský. 2012. Prague dependency style treebank for Tamil. In Proceedings of Eighth International Conference on Language Resources and Evaluation (LREC 2012), pages 1888–1894, İstanbul, Turkey. find
  24. Vinit Ravishankar. 2017. A Universal Dependencies treebank for Marathi. In Proceedings of the 16th International Workshop on Treebanks and Linguistic Theories, pages 190–200, Prague, Czech Republic. find
  25. Simran Khanuja, Diksha Bansal, Sarvesh Mehtani, Savya Khosla, Atreyee Dey, Balaji Gopalan, Dilip Kumar Margam, Pooja Aggarwal, Rajiv Teja Nagipogu, Shachi Dave, Shruti Gupta, Subhash Chandra Bose Gali, Vish Subramanian, and MuRIL: Multilingual Partha Talukdar. 2021. representations for Indian langua… find our card
  26. Radim Rehurek and Petr Sojka. 2011. Gensim–python framework for vector space modelling. NLP Centre, Faculty of Informatics, Masaryk University, Brno, Czech Republic, 3(2). find
  27. Christopher Manning, Mihai Surdeanu, John Bauer, Jenny Finkel, Steven Bethard, and David McClosky. 2014. The Stanford CoreNLP natural language processing toolkit. In Proceedings of 52nd Annual Meeting of the Association for Computational Linguistics: System Demonstrations, pages 55–60, Baltimore, Ma… find our card
  28. Tomas Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Efficient estimation of word representations in vector space. Preprint, arXiv:1301.3781. find
  29. Marie Mikulová, Milan Straka, Jan Štěpánek, Barbora Štěpánková, and Jan Hajic. 2022. Quality and efficiency of manual annotation: Pre-annotation bias. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 2909–2918, Marseille, France. European Language Resources Associ… find
  30. Bharti Nathani, Nisheeth Joshi, and G.N. Purohit. 2020. Design and development of unsupervised stemmer for Sindhi language. Procedia Computer Science, 167:1920–1927. International Conference on Computational Intelligence and Data Science. find
  31. Natalia Silveira, Timothy Dozat, Marie-Catherine de Marneffe, Samuel Bowman, Miriam Connor, John Bauer, and Christopher D. Manning. 2014. A gold standard dependency corpus for English. In Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC-2014). find
  32. Milan Straka. 2018. UDPipe 2.0 prototype at CoNLL 2018 UD shared task. In Proceedings of the CoNLL 2018 Shared Task: Multilingual Parsing from Raw Text to Universal Dependencies, pages 197–207, Brussels, Belgium. Association for Computational Linguistics. find our card
  33. Amir Zeldes. 2017. The GUM corpus: Creating multilayer resources in the classroom. Language Resources and Evaluation, 51(3):581–612. find
  34. A Features Table 6 demonstrates the morphological features used for the various verb forms described in section 4.6. B Postposition Examples Beryne Odeny and Raffaella Bosurgi. 2022. Time to end parachute science. PLOS Medicine, 19(9):1–3. find
  35. Martha Palmer, Rajesh Bhatt, Bhuvana Narasimhan, Owen Rambow, Dipti Misra Sharma, and Fei Xia. 2009. Hindi syntax: Annotating dependency, lexical predicate-argument structure, and phrase structure. In The 7th International Conference on Natural Language Processing, pages 14–17. find
  36. As explained in section 4.4, genitive adpositions in Sindhi inflect to agree with the associated noun, whereas ADPs representing location do not inflect. For example, consider the locative ADPs in these sentences: Jeffrey Pennington, Richard Socher, and Christopher Manning. 2014. GloVe: Global vecto… find
  37. Peng Qi, Yuhao Zhang, Yuhui Zhang, Jason Bolton, and Christopher D. Manning. 2020. Stanza: A Python • “ ﻲﭘﻮٽ ﻲﺗ ﻲﭥﻣ” (mathe te topī, “A cap on the head”) the noun “ ﻮﭥﻣ” (matho, “head”) shifts to its oblique form “ ﻲﭥﻣ” (mathe) before “ ﻲﺗ” (te). find
  38. “ نﺎڪد ۾ يڙﺎﭘ” (pāṛe mē dukān, “A shop in the neighborhood”) “ وڙﺎﭘ” (pāṛo, “neighborhood”) becomes “ يڙﺎﭘ” (pāṛe). find

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/bauer-2025-universal-dependencies-sindhi/