Database Paper Browser

Back to papers

Visual Template Inference for Data Extraction from Documents

Summary: TWIX infers latent visual templates for programmatically generated documents by clustering consistently co-located fields and enforcing alignment constraints (e.g., column/header and key/value alignment) to assemble templates for extraction. Template-driven extraction achieves >25% higher precision/recall than Evaporate/Textract/Azure/GPT-4-Vision on 34 datasets and is massively more scalable (≈520× faster, ≈3,786× cheaper) on large corpora. (summarized by gpt-5-mini on Feb 11 2026)

Paper ID
7436
Venue
SIGMOD
Year
2026
Pagerank
4.1945683e-05
Overall Rank
10,126 | 29.56%
DOI
10.1145/3769840

Incoming Non-self Citations Over Time

No non-self incoming citations found for this paper in this database.

Authors

Incoming Citations (Sorted by Pagerank)

Showing 0 of 0 citing papers.

Rank Citing Paper Year Venue Pagerank
Previous Page 1 / 1 Next

Outgoing Citations (Sorted by Pagerank)

Showing 15 of 15 cited papers.

Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.

Previous Page 1 / 1 Next

Semantically Similar Papers