Elamur Research⌕
☰
home / papers / 2026 / TEN: Table Explicitization, Neurosymbolically

TEN: Table Explicitization, Neurosymbolically

Paper · 2026 · Nikita Mehrotra, Aayush Kumar, Sumit Gulwani et al.

Summary

The paper addresses "table explicitization": recovering a clean table from plain text that has lost its row and column delimiters, such as content copy-pasted from PDFs or emails, or flattened by OCR. Layout-based tools fail on this kind of input. LLMs can handle it, but they are non-deterministic and may hallucinate cells or produce inconsistent structures across runs. The authors, from Microsoft, propose TEN, a neurosymbolic system built on a three-part loop:

  • A generator LLM drafts a table using Structural Decomposition prompting, which locates regions and boundaries before building the table.
  • A deterministic rule-based checker flags structural and content violations.
  • A critique LLM converts those violations into targeted repair instructions.

The loop repeats until all constraints are met. Evaluation covers FinTabNet, PubTabNet, BrokenCSV, WikiTables and a small FinReCon20 set. TEN beats neural baselines on exact match and hallucination rates. In a user study, participants rated its tables as more accurate. TEN is already used as an internal enterprise tool, which suggests that symbolic validation makes LLM-based extraction reliable enough for production.

Key points

  • Symbolic checker rules: entity-consistency checks (dates, URLs, emails and similar), column signatures, merged-cell detection, wrong-delimiter fragmentation, bracket matching and empty-row detection. The checker also scores goodness, badness, coverage and hallucination, and runs in under 50 ms on a CPU.
  • Ablation: combining symbolic and LLM feedback gives the best exact match on every dataset:
    • 36.80 on Wikipedia
    • 81.6 on BrokenCSV
    • 61.60 on PubTabNet
    • 21.73 on FinTabNet
  • User study (21 participants): TEN's tables received higher accuracy ratings (5.0 vs. 4.3) and were preferred in over 60% of pairwise comparisons. However, the effort needed to verify or fix the tables did not differ significantly.
  • Cost and speed: roughly $0.08–$0.15 per table with GPT-4 over 2–3 iterations. Latency is 5–15 s per table, reduced to a median of 3–6 s through early stopping and caching.
  • Datasets: four benchmarks of 1000 tables each, plus FinReCon20 with 20 tables.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🏢MicrosoftRedmond · United States325 works · 142,931 citations

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 works

Cite

Cite as Nikita Mehrotra, Aayush Kumar, Sumit Gulwani, Arjun Radhakrishna and Ashish Tiwari. 2026. TEN: Table Explicitization, Neurosymbolically. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-industry.138

@inproceedings{mehrotra2026ten,
  title = {{TEN: Table Explicitization, Neurosymbolically}},
  author = {Nikita Mehrotra, Aayush Kumar, Sumit Gulwani, Arjun Radhakrishna and Ashish Tiwari},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 6: Industry Track)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-industry.138},
  url = {https://doi.org/10.18653/v1/2026.acl-industry.138}
}

References

35 references, 6 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Ewa Andrejczuk, Julian Martin Eisenschlos, Francesco Piccinno, Syrine Krichene, and Yasemin Altun. 2022. Table-to-text generation and pre-training with tabt5. Preprint, arXiv:2210.09162. find
  2. Manuel Aristáran, Mike Tigas, Jeremy B. Merrill, and Jason Das. 2018. Tabula: Extract tables from pdfs. Open-source tool, https://tabula.technology/. ↗
  3. Kimberly A. Barchard and Larry A. Pace. 2011. Preventing human error: The impact of data entry methods on data accuracy and statistical results. Computers in Human Behavior, 27(5):1834–1839. 2009 Fifth International Conference on Intelligent Computing. find
  4. Lucas Beyer, Andreas Steiner, André Susano Pinto, Alexander Kolesnikov, Xiao Wang, Daniel Salz, Maxim Neumann, Ibrahim Alabdulmohsin, Michael Tschannen, Emanuele Bugliarello, Thomas Unterthiner, Daniel Keysers, Skanda Koppula, Fangyu Liu, Adam Grycner, Alexey Gritsenko, Neil Houlsby, Manoj Kumar, Ke… find
  5. Chandra Sekhar Bhagavatula, Thanapon Noraset, and Doug Downey. 2013. Methods for exploring and mining tables on wikipedia. In Proceedings of the ACM SIGKDD Workshop on Interactive Data Exploration and Analytics, IDEA ’13, page 18–26, New York, NY, USA. Association for Computing Machinery. find
  6. Juan Carlos Crisanto, Cris Benson Leuterio, Jermy Pre-nio, and Jeffery Yong. 2024. Regulating ai in the financial sector: recent developments and main challenges. Fsi insights on policy implementation; no. 63, Financial Stability Institute, Bank for International Settlements. find
  7. Xiang Deng, Huan Sun, Alyssa Lees, You Wu, and Cong Yu. 2020. TURL: table understanding through representation learning. CoRR, abs/2006.14806. — [arXiv:2006.14806](https://arxiv.org/abs/2006.14806) find
  8. Katherine Hermann, Sean Welleck, Amir Yazdanbakhsh, and Peter Clark. 2023. Self-refine: Iterative refinement with self-feedback. In Advances in Neural Information Processing Systems, volume 36, pages 46534–46594. Curran Associates, Inc. find
  9. Tim Mathes, Pauline Klaßen, and Dawid Pieper. 2017. Frequency of data extraction errors and methods to increase data extraction quality: a methodological review. BMC Medical Research Methodology, 17(1):152. find
  10. Quinn McNemar. 1947. Note on the sampling error of the difference between correlated proportions or percentages. Psychometrika, 12(2):153–157. find
  11. Shubham Singh Paliwal, Vishwanath D, Rohit Rahul, Monika Sharma, and Lovekesh Vig. 2019. Tablenet: Deep learning model for end-to-end table detection and tabular data extraction from scanned document images. In 2019 International Conference on Document Analysis and Recognition (ICDAR), pages 128– 13… find
  12. Mohammad Raza and Sumit Gulwani. 2017. Automated Data Extraction Using Predictive Program Synthesis. In AAAI, pages 882–890. find
  13. Zheye Deng, Chunkit Chan, Weiqi Wang, Yuxi Sun, Wei Fan, Tianshi Zheng, Yauwai Yim, and Yangqiu Song. 2024. Text-tuple-table: Towards information integration in text-to-table generation via global tuple extraction. Preprint, arXiv:2404.14215. find
  14. Mohammad Raza and Sumit Gulwani. 2020. Web data extraction using hybrid program synthesis: A combination of top-down and bottom-up inference. In Proceedings of the 2020 ACM SIGMOD International Conference on Management of Data, pages 1967– 1978. find
  15. Haoyu Dong, Shijie Liu, Zhouyu Fu, Shi Han, and Dongmei Zhang. 2019. Semantic structure extraction for spreadsheet tables with a multi-task learning architecture. In Workshop on Document Intelligence at NeurIPS 2019. find
  16. Jeff Sauro and Joseph S. Dumas. 2009. Comparison of three one-question, post-task usability questionnaires. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, CHI ’09, page 1599–1608, New York, NY, USA. Association for Computing Machinery. find
  17. Zhibin Gou, Zhihong Shao, Yeyun Gong, Yelong Shen, Yujiu Yang, Nan Duan, and Weizhu Chen. 2024. Critic: Large language models can self-correct with tool-interactive critiquing. Preprint, arXiv:2305.11738.
  18. Jonathan Herzig, Pawel Krzysztof Nowak, Thomas Müller, Francesco Piccinno, and Julian Martin Eisenschlos. 2020. TAPAS: weakly supervised table parsing via pre-training. CoRR, abs/2004.02349.
  19. Minsoo Khang and Teakgyu Hong. 2025. Tflop: Table structure recognition framework with layout pointer mechanism. Preprint, arXiv:2501.11800. — [arXiv:2501.11800](https://arxiv.org/abs/2501.11800) find
  20. Minghao Li, Lei Cui, Shaohan Huang, Furu Wei, Ming Zhou, and Zhoujun Li. 2019. Tablebank: Table benchmark for image-based table detection and recognition. CoRR, abs/1903.01949. find
  21. Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, Shashank Gupta, Bodhisattwa Prasad Majumder, Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik R Narasimhan, and Shunyu Yao. 2023. Reflexion: language ag…
  22. Yusuke Shinyama. 2014. Pdfminer: Python pdf parser and analyzer. https://github.com/euske/ pdfminer. Accessed: 2025-04-17. ↗
  23. Mukul Singh, Gust Verbruggen, Vu Le, and Sumit Gulwani. 2024. Tabularis revilio: Converting text to tables. In Proceedings of the 33rd ACM International Conference on Information and Knowledge Management, CIKM ’24, page 4056–4060, New York, NY, USA. Association for Computing Machinery. find
  24. Ray Smith. 2007. An overview of the tesseract ocr engine. In Proceedings of the Ninth International Conference on Document Analysis and Recognition (ICDAR), pages 629–633. IEEE. find
  25. Araz Taeihagh. 2025. Governance of generative ai. Policy and Society, 44(1):1–22. find
  26. G. J. J. van den Burg, A. Nazábal, and C. Sutton. 2019. Wrangling messy csv files by detecting row and type patterns. Data Min. Knowl. Discov., 33(6):1799–1820. find
  27. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed H. Chi, Quoc Le, and Denny Zhou. 2022. Chain of thought prompting elicits reasoning in large language models. CoRR, abs/2201.11903.
  28. Xueqing Wu, Jiacheng Zhang, and Hang Li. 2022. Text-to-table: A new way of information extraction. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 2518–2533, Dublin, Ireland. Association for Computational Linguistics. find
  29. Wenda Xu, Daniel Deutsch, Mara Finkelstein, Juraj Juraska, Biao Zhang, Zhongtao Liu, William Yang Wang, Lei Li, and Markus Freitag. 2024. LLMRefine: Pinpointing and refining large language models via fine-grained actionable feedback. In Findings of the Association for Computational Linguistics: NAAC… find
  30. Yang Xu, Yiheng Xu, Tengchao Lv, Lei Cui, Furu Wei, Guoxin Wang, Yijuan Lu, Dinei Florencio, Cha Zhang, Wanxiang Che, Min Zhang, and Lidong Zhou. 2021. LayoutLMv2: Multi-modal pre-training for visually-rich document understanding. In Proceedings of the 59th Annual Meeting of the Association for Comp…
  31. Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Thomas L. Griffiths, Yuan Cao, and Karthik R Narasimhan. 2023. Tree of thoughts: Deliberate problem solving with large language models. In Thirty-seventh Conference on Neural Information Processing Systems. find
  32. Pengcheng Yin, Graham Neubig, Wen-tau Yih, and Sebastian Riedel. 2020. Tabert: Pretraining for joint understanding of textual and tabular data. CoRR, abs/2005.08314.
  33. Kaizhong Zhang and Dennis Shasha. 1989. Simple fast algorithms for the editing distance between trees and related problems. SIAM Journal on Computing, 18(6):1245–1262. find
  34. Xinyi Zheng, Douglas Burdick, Lucian Popa, and Nancy Xin Ru Wang. 2020. Global table extractor (GTE): A framework for joint table identification and cell structure recognition using visual context. CoRR, abs/2005.00589. — [arXiv:2005.00589](https://arxiv.org/abs/2005.00589) find
  35. Xu Zhong, Elaheh ShafieiBavani, and Antonio Jimeno Yepes. 2019. Image-based table recognition: data, model, and evaluation. arXiv preprint arXiv:1911.10683. — [arXiv:1911.10683](https://arxiv.org/abs/1911.10683) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/mehrotra-2026-ten-table-explicitization-neurosymbolically/