Encoding Logical Relations of Chinese Complex Sentences within the Universal Dependencies Framework

Summary

Clauses inside Chinese complex sentences stand in logical relations such as cause, concession or conjunction, which are often left unmarked by connectives and can nest hierarchically. Standard Universal Dependencies flattens all of this into generic inter-clause labels like advcl, ccomp or parataxis. The authors replace those generic labels with 13 finer logical-relation types while leaving every other dependency untouched, so the resulting trees stay structurally compatible with UD and clause hierarchy plus relation type become a single parsing problem. They annotate about 1,769 newswire sentences for training and re-annotate the GSD-simp test split as an evaluation set. A BERT-based biaffine parser and a fine-tuned Qwen-3 model are compared with several general-purpose LLMs (GPT-4o, GPT-5, Claude 4, DeepSeek V3.2). The fine-tuned Qwen-3-8B scores best with UAS/LAS 0.840/0.757. The study shows that inter-clause logic fits into the familiar UD format, but implicit and deeply nested relations remain hard for automatic parsers, and accuracy degrades as sentences contain more clauses.

How Mova uses it

• Typical failures of LLM annotators for UD — a seed (a short declarative rule note) listing typical failure modes of LLM-based UD annotators, with a check for each. This paper is one of the sources for the failure mode "sentence complexity: accuracy falls as the number of clauses grows".
• The observation feeds the general design of the annotation pipeline: the deterministic Rust parser annotates first and a large model only edits the draft, and evaluation always uses an independent gold standard rather than corrected model output.
• The 13-relation scheme itself is not implemented; Mova's English pipeline keeps plain UD inter-clause labels.

Effectiveness in Mova

Background reading: the paper supports one item of a checklist of LLM annotator errors. Not measured separately.

👨‍🔬💥

Topics: Universal Dependencies

Cite

The paper itself, not this card. BibTeX from the library's record of the work. DOI 10.63317/2uasimdgfqin

@inproceedings{zhu2026encoding,
  title = {{Encoding Logical Relations of Chinese Complex Sentences within the Universal Dependencies Framework}},
  author = {Zhu, Hongpu and Xu, Hongzhi},
  booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference},
  year = {2026},
  doi = {10.63317/2uasimdgfqin},
  url = {https://doi.org/10.63317/2uasimdgfqin}
}

References

30 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Shuyuan Cao, Nianwen Xue, Iria da Cunha, Mikel Iruskieta, and Chuan Wang. 2017. Discourse Segmentation for Building a RST Chinese Treebank. In Proceedings of the 6th Workshop on Recent Advances in RST and Related Formalisms, pages 73–81, Santiago de Compostela, Spain. Association for Computational L… find
2. Yi Cheng and Sujian Li. 2019. Zero-shot chinese discourse dependency parsing via cross-lingual mapping. arXiv preprint arXiv:1911.12014. find
3. Timothy Dozat and Christopher D. Manning. 2017. Deep Biaffine Attention for Neural Dependency Parsing. ArXiv:1611.01734 [cs]. find our card
4. Yunlong Fan, Bin Li, Yikemaiti Sataer, Miao Gao, Chuanqi Shi, Siyi Cao, and Zhiqiang Gao. 2023. Hierarchical Clause Annotation: Building a Clause-Level Corpus for Semantic Parsing with Complex Sentences. Applied Sciences, 13(16):9412. Number: 16 Publisher: Multidisciplinary Digital Publishing Instit… find
5. Han He and Jinho D. Choi. 2021. The Stem Cell Hypothesis: Dilemma behind Multi-Task Learning with Transformer Encoders. ArXiv:2109.06939 [cs]. find
6. Chu-Ren Huang and Dingxu Shi. 2016. A reference grammar of Chinese. Cambridge University Press. find
7. Feng Jiang, Sheng Xu, Xiaomin Chu, Peifeng Li, Qiaoming Zhu, and Guodong Zhou. 2018. Mcdtb: a macro-level chinese discourse treebank. In Proceedings of the 27th International Conference on Computational Linguistics, pages 3493–3504. find
8. Dan Kondratyuk and Milan Straka. 2019. 75 Languages, 1 Model: Parsing Universal Dependencies Universally. ArXiv:1904.02099 [cs]. find our card
9. Alex Lascarides and Nicholas Asher. 2008. Segmented Discourse Representation Theory: Dynamic Semantics With Discourse Structure. In Harry Bunt and Reinhard Muskens, editors, Computing Meaning, pages 87–124. Springer Netherlands, Dordrecht. find
10. Herman Leung, Rafaël Poiret, Tak-sum Wong, Xinying Chen, Kim Gerdes, and John Lee. 2016. Developing Universal Dependencies for Mandarin Chinese. In Proceedings of the 12th Workshop on Asian Language Resources (ALR12), pages 20–29, Osaka, Japan. find
11. The COLING 2016 Organizing Committee. find
12. Bin Li, Miao Gao, Yunlong Fan, Yikemaiti Sataer, Zhiqiang Gao, and Yaocheng Gui. 2022. DynGL-SDP: Dynamic graph learning for semantic dependency parsing. In Proceedings of the 29th International Conference on Computational Linguistics, pages 3994–4004. find
13. Yunhua Liang. 2002. An analysis of the compact complex structure in modern chinese. 19(2):99– 106. find
14. William C. Mann and Sandra A. Thompson. 1988. Rhetorical Structure Theory: Toward a functional theory of text organization. Text - Interdisciplinary Journal for the Study of Discourse, 8(3). find our card
15. Daniel Marcu. 2000. The theory and practice of discourse parsing and summarization. MIT press. find
16. Siyao Peng, Yang Janet Liu, and Amir Zeldes. 2022. GCDT: A Chinese RST Treebank for Multigenre and Multilingual Discourse Parsing. ArXiv:2210.10449 [cs]. find
17. Rafaël Poiret, Tak-Sum Wong, John Lee, Kim Gerdes, and Herman Leung. 2023. Universal Dependencies for Mandarin Chinese. Language Resources and Evaluation, 57(2):673–710. find
18. Rashmi Prasad, Bonnie Webber, and Aravind Joshi. 2017. The Penn Discourse Treebank: An Annotated Corpus of Discourse Relations. In Nancy Ide and James Pustejovsky, editors, Handbook of Linguistic Annotation, pages 1197–1217. Springer Netherlands, Dordrecht. find
19. Livio Robaldo. 2008. The penn discourse treebank 2.0. In Lrec. find our card
20. Kaili Sun, Yuan Li, Dunhua Deng, and Yang Li. 2019. Multi-channel CNN based inner-attention for compound sentence relation classification. IEEE Access, 7:141801–141809. Publisher: IEEE. find
21. Wenhong Tian, Yinquan Gao, Houwen Huang, Zaiwan Li, and Zhaoyang Zhang. 2019. Implicit Discourse Relation Analysis Based on Multi-task Bi-LSTM. Journal of Chinese Information Processing, 33(5):47–53. find
22. Nianwen Xue and Yaqin Yang. 2011. Chinese sentence segmentation as comma classification. In Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies, pages 631–635, Portland, Oregon, USA. Association for Computational Linguistics. find
23. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Yang, Jiaxi Yang, Jing Zhou, Jin… find
24. Jincai Yang, Zhongzhong Chen, Xianjun Shen, and Jinzhu Hu. 2017. Automatic recognition of relation category of non-saturated compound sentences with two clauses. Application Research of Computers/Jisuanji Yingyong Yanjiu, 34(10). find
25. Jingcai Yang, Yuxin Cao, Hu Quan, and Cai Xu-Xun. 2022. Automatic Recognition of Chinese Compound Sentence Relation Based on BERT-FHAN Model and Sentence Features. Computer Systems & Applications, 31(9):233–240. find
26. Lanjun Zhou, Binyang Li, Zhongyu Wei, and Kam-Fai Wong. 2014. The CUHK Discourse Tree-Bank for Chinese: Annotating Explicit Discourse Connectives for the Chinese TreeBank. In LREC, pages 942–949. find
27. Yuping Zhou and Nianwen Xue. 2012. PDTB-style discourse annotation of Chinese text. In Proceedings of the 50th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 69–77. find
28. Yuping Zhou and Nianwen Xue. 2015. The Chinese Discourse TreeBank: a Chinese corpus annotated with discourse relations. Language Resources and Evaluation, 49(2):397–431. 10. find
29. Language Resource References Parker, Robert, Graff, David, Chen, Ke, Kong, Junbo, and Maeda, Kazuaki. 2011. Chinese Gigaword Fifth Edition. find
30. Qi, Peng and Yasuoka, Koichi. 2022. UD Chinese GSDSimp. PID https://github.com/UniversalDependencies. ↗

This page in Markdown → https://md.mova.elamur.ai/free/en/docs/papers/zhu-2026-encoding-logical-relations-chinese-complex/