TEN: Table Explicitization, Neurosymbolically

2025-08-12

AI Summary

Table extraction from semi-structured text lacking consistent delimiters is highly susceptible to hallucinations in large language models (LLMs) and struggles to guarantee structural compliance. Method: We propose TEN, a neuro-symbolic collaborative framework that first employs structure-decomposition prompting to guide an LLM in generating an initial table, then deploys a lightweight symbolic validator to detect structural errors—including row-column alignment violations and type inconsistencies—and finally drives iterative self-correction via critical feedback. Contribution/Results: TEN introduces the first “prompt–validate–revise” closed-loop mechanism, achieving breakthroughs in hallucination suppression and verifiable structural correctness. Experiments demonstrate significant improvements over purely neural baselines across multiple benchmarks, with marked gains in exact-match accuracy. A user study confirms that TEN’s outputs are more accurate and easier to verify, with over 60% of cases preferred by human evaluators.

Technology Category

• Natural Language Processing: Information Extraction
• Cognitive Modeling & Cognitive Systems: Symbolic Representations
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

We present a neurosymbolic approach, TEN, for extracting tabular data from semistructured input text. This task is particularly challenging for text input that does not use special delimiters consistently to separate columns and rows. Purely neural approaches perform poorly due to hallucinations and their inability to enforce hard constraints. TEN uses Structural Decomposition prompting - a specialized chain-of-thought prompting approach - on a large language model (LLM) to generate an initial table, and thereafter uses a symbolic checker to evaluate not only the well-formedness of that table, but also detect cases of hallucinations or forgetting. The output of the symbolic checker is processed by a critique-LLM to generate guidance for fixing the table, which is presented to the original LLM in a self-debug loop. Our extensive experiments demonstrate that TEN significantly outperforms purely neural baselines across multiple datasets and metrics, achieving significantly higher exact match accuracy and substantially reduced hallucination rates. A 21-participant user study further confirms that TEN's tables are rated significantly more accurate (mean score: 5.0 vs 4.3; p = 0.021), and are consistently preferred for ease of verification and correction, with participants favoring our method in over 60% of the cases.

Problem

• Extracting tabular data from inconsistent delimiters text
• Reducing hallucinations in neural table extraction methods
• Ensuring well-formed tables with symbolic constraint checking

Innovation

• Neurosymbolic approach combining LLM and symbolic checker
• Structural Decomposition prompting for initial table generation
• Self-debug loop with critique-LLM for error correction