Sign in Subscribe
Electronic records

Breakthrough Patent Application for Table Extraction from Unstructured Documents

Strategic Insights

— 1 min read

A groundbreaking patent application by inventors Stephan Schwiebert and Velislava Petrova Yanchina has made available online, detailing a revolutionary method for extracting tables from unstructured documents. The application, filed on April 4, 2025, was posted on October 16, 2025. The inventors' work aims to tackle the long-standing challenge of identifying tables from unstructured text, particularly in documents such as PDFs, which are difficult to edit and pose significant challenges for automated table detection.

Key Takeaways:

  • The inventors propose a computer-implemented method for accessing data defining a plurality of glyphs, generating document layout data, and identifying contents of cells of at least one document table.
  • The method generates an output based on the identified contents of cells of at least one document table, including creating an editable document table corresponding to the at least one document table.
  • The editable document table includes text with an alignment, and the alignment of the text is determined based on the text block of the plurality of text blocks that has corresponding content and one or more of the text block links for that text block.
  • The inventors describe a computer system configured to perform the methods, as well as a non-transitory computer-readable memory comprising instructions to cause a computer processor to perform the methods.
  • The patent application includes 22 claims, detailing the scope of the invention and the various embodiments of the method and system.
  • The inventors' work builds upon previous research in the field, aiming to improve the accuracy and efficiency of table detection from unstructured documents.

Statistics:

  • The patent application was filed on April 4, 2025, and posted online on October 16, 2025.
  • The inventors' method comprises 22 claims and is configured to perform automated table detection from unstructured documents such as PDFs.
  • At least one document table corresponds to one table grid in the proposed method.

Sources:

  • "Systems and methods for extracting tables from documents" by Stephan Schwiebert and Velislava Petrova Yanchina, U.S. Patent Application Number 20250322148, filed April 4, 2025, and posted October 16, 2025. Patent URL: https://ppubs.uspto.gov/pubwebapp/external.html?q=(20250322148)&db=US-PGPUB&type=ids

Read more