Back to papers

Making Table Understanding Work in Practice

Summary: Identifies a gap between high-accuracy DL table-understanding models on benchmarks and real-world deployments that still rely on heuristics/regex for a few semantic types. Characterizes practical needs—robustness, coverage, explainability, efficiency, calibration, domain adaptation—and prescribes evaluation and design directions to make table understanding usable in production. (summarized by gpt-5-mini on Feb 09 2026)

Paper ID

434

Venue

CIDR

Year

2022

Pagerank

4.4229886e-05

Overall Rank

8,915 | 38.05%

DOI

-